Jud Be Hyung

HOW JUDGES DECIDE: A MULTIDIMENSIONAL EMPIRICAL TYPOLOGY OF JUDICIAL STYLES IN THE

FEDERAL COURTS

Corey Rayburn Yung

ABSTRACT

This article calls into question fundamental premises related to the study of judicial decisionmaking by political science and law scholars. The three dominant explanations for how judges make decisions (the attitudinal, strategic, and legal models) have each relied on three highly questionable assumptions: 1) Homogeneity – judges should be studied as a monolithic group that utilizes similar, if not identical, approaches to judging; 2) Unidimensionality – judges decide cases based upon a single goal that should be measured along a single dimension; 3) Isolation – each level of judges should be gauged independently of other judges.

In the alternative, this article offers a typological approach which recognizes judicial heterogeneity, multidimensional behavior, and interconnectedness among judges at different levels within the judiciary. The study utilizes new data of over 10,000 cases from eleven Courts of Appeals in 2008 to create measures for judicial activism, ideology, independence, and partisanship. Based upon those four metrics, statistical cluster analysis is used to identify nine distinct judging styles: Trailblazing, Consensus Building, Stalwart, Regulating, Steadfast, Collegial, Pragmatic, Minimalistic, Error Correcting. These types provide a much more nuanced and fuller account of how judges actually decide cases.

Further, the typological model is tested against traditional ideology-based explanations of judicial behavior using a separate new dataset of over 25,000 cases from the same eleven circuits in 2009. Notably, the judicial types prove substantially better in forecasting decisions by judges. The increased ability to predict dissents is particularly significant as the typology model is accurate in 72 to 90% of cases. In contrast, the ideology-based model shows just modest improvement over random guesses. The study, using regression analysis, also identifies statistically significant relationships between background traits and the nine judicial types. Among other findings, ranking of law school attended, circuit, and senior status all are correlated with a judge’s style type. The article concludes with ideas for new directions for empirical research that move beyond the stagnant models that have reigned supreme in recent studies.

2 HOW JUDGES DECIDE [9-Feb-11

HOW JUDGES DECIDE: A MULTIDIMENSIONAL EMPIRICAL TYPOLOGY OF JUDICIAL STYLES IN THE

FEDERAL COURTS

Corey Rayburn Yung*

INTRODUCTION Style is the perfection of a point of view.

– Richard Eberhart1

The dominant models of judicial2

decisionmaking rely on three dubious assumptions that have not been empirically tested and validated:

1) Homogeneity – judges should be studied as a monolithic group that utilizes similar, if not identical, approaches to judging.3

2) Unidimensionality – judges decide cases based upon [politics, strategy, law] and [politics, strategy, law] should be measured along a single dimension.

4

* Associate Professor of Law at John Marshall Law School. I would like to thank

David L. Schwartz, Ramsey Donnell, Raizel Liebler, and Stuart Ford for their comments, support, and suggestions. For the tireless efforts in helping to create the datasets used in this study, I would like to thank Ashley Ahn, Christopher Hack, Dan Callandriello, Diane O'Connell, Gina Lencioni, Jacklyn Martin, Jessica Kull, Jessica Thornhill, Jillian Eckhart, Kim Regan, Lydia San, Michael Carriglio, Miguel Larios, Molly Glanz, Patrycja Rynduch, Rajesh Mathew, Robert Breslin, Sandra Esposito, and Stephanie Lieberman. Lastly, I want to thank John Corkery, Ralph Ruebner, and The John Marshall Law School without whose incredible support, this project would not have been possible.

1 RICHARD EBERHART, THE QUARRY 111 (1964). 2 “Judges” and “judicial” in this article are meant to specifically exclude the Justices of

the Supreme Court of the United States. While some of the criticisms of the dominant approaches in this article apply to Supreme Court studies, far more research of the Court has been done that meets the challenges outlined herein. Indeed, in some cases, the excellent empirical research of the Justices has inspired some of the measures described and utilized in this Article. See, e.g., Andrew D. Martin & Kevin M. Quinn, Dynamic Ideal Point Estimation via Markov Chain Monte Carlo for the U.S. Supreme Court, 1953-1999, 10 POL. ANALYSIS 134, 139-40 (2002).

3 See, e.g., Daniel Klerman, Jurisdictional Competition and the Evolution of the Common Law, 74 U. CHI. L. REV. 1179, 1203 (1997 (“The jurisdictional competition model assumes that judges are homogeneous. Of course, they were not. Introducing heterogeneity would complicate the model without substantially altering the conclusions.”).

4 This assumption has been most often made in studies of judicial ideology. Carolyn Shapiro, Coding Complexity: Bringing Law to the Empirical Analysis of the Supreme Court, 60 HASTINGS L. J. 477, 501 (2009) (“These scholars assume that in deciding a case, the Justices need only consider their preferences about a single question and that the

9-Feb-11] HOW JUDGES DECIDE 3

3) Isolation – each judge or level of judges should be gauged independently of other actors in the judicial system.5

Each of those premises has undergirded the most common explanations of judicial behavior used by law and political science scholars: the attitudinal (political), strategic, and formal (legal) models.6 Research addressing the common assumptions has been limited, in part, because of the available data – there simply has not been a publicly accessible dataset that could be effectively used to study relationships between diverse judges while concurrently utilizing multiple behavioral metrics.7

Justices’ preferences can be arrayed along a single ideological dimension.”) (internal quotation marks omitted). Although some scholarship has tested and validated the unidimensionality assumption at the Supreme Court level, see, Tonja Jacobi & Matthew Sag, Taking the Measure of Ideology: Empirically Measuring Supreme Court Cases, 98 GEO. L.J. 1, 8 (2009) (“…there is considerable evidence that a single dimension captures the vast majority of judicial behavior. The fact that one dimension captures most judicial behavior does not imply that legal analysis is simplistic, only that most judicial considerations are generally quite highly correlated.”), no similar studies have been done for the Courts of Appeals. See, Joshua B. Fischman & David S. Law, What is Judicial Ideology, and How Should We Measure It?, 29 WASH. U. J.L. & POL'Y 133, 168-170 (2009) (“At the same time, the assumption of unidimensional ideology may not hold as well for other courts as it does for the United States Supreme Court. Very little is known about the dimensionality of ideology on other courts, and further investigation is surely warranted.”).

However, the

5 A notable exception to this general rule has been the work regarding panel effects. See, e.g., Frank B. Cross & Emerson H. Tiller, Judicial Partisanship and Obedience to Legal Doctrine: Whistleblowing on the Federal Courts of Appeals, 107 YALE L.J. 2155, 2168-75 (1998); Richard L. Revesz, Environmental Regulation, Ideology, and the D.C. Circuit, 83 VA. L. REV. 1717, 1751-56 (1997); Pauline T. Kim, Deliberation and Strategy on the United States Courts of Appeals: An Empirical Exploration of Panel Effects, 157 U. PA. L. REV. 1319 (2009). However, even within the very valuable scholarship regarding panel effects, research has been limited to judges operating on the same panels. There has not been an exploration of interrelations between judges working at different levels in the judicial system. This omission is particular notable given the rigid hierarchal structure of the judicial systems in the United States wherein review of lower courts is the primary function of higher courts.

6 Frank B. Cross, Decisionmaking in the U.S. Circuit Courts of Appeals, 91 CALIF. L. REV. 1457, 1461–90 (2003) [hereinafter “Cross Decisionmaking”]. Cross also discussed a fourth model for judicial decisionmaking: the litigant-driven theory. Id. at 1490. However, the litigant-driven theory has not been widely studied and, as Cross noted, evidence that litigants drive decision-making in the Courts of Appeals has been very limited. Id. at 1514 (“The data do not provide strong support for the economic litigant-driven hypothesis regarding judicial decisionmaking.”).

7 Previous empirical research concerning federal appellate judges has relied primarily upon the United States Courts of Appeals Database (“Songer Database”), which includes more than 18,000 opinions from 1925 to 1996. ASHLYN K. KUERSTEN & DONALD R. SONGER, DECISIONS ON THE U.S. COURTS OF APPEALS 1 (2001).


assumptions of homogenous judicial decisionmaking, unidimensional measuring options, and limited interconnectedness have gained scholarly momentum beyond simple data availability concerns. The result has been that much of the empirical scholarship about law and courts has focused on variations of the same basic models and theories using the untested hypotheses related to a monolithic judiciary.

This study drew from recent scholarship and new data to question the three untested assumptions of the major theorized global models. In particular, the study described herein integrated trial and appellate decisionmaking into one system yielding behavioral metrics of the traits associated with judicial activism, independence, partisanship, and ideology.8 Further, applying statistical cluster analysis to construct a typology of judges, the study found an incredible diversity of judicial behavior along multiple dimensions. Primarily, the study identified nine distinctive decisionmaking styles of judges on the United States Courts of Appeals: Trailblazing, Consensus Building, Stalwart, Regulating, Steadfast, Collegial, Pragmatic, Minimalistic, and Error Correcting. These varied approaches provided a fuller picture of the judicial process that recognized the differing motivations, goals, and concerns of judges. Instead of assuming all judges behaved in a similar manner, this study found substantial variation wherein ideology, the most popular explanation for judicial decisionmaking,9

After completion of the typology, the study tested the judicial styles against traditional ideology-based models in predicting future judicial votes. Most notably, the two most basic dispositional choices an appellate judge could make, to dissent or reverse, were both better explained using the judge style categories. In the case of predicting the dissenting judge, the classic ideology-based model showed just modest improvements over predictions based upon a random guess with model consistency levels of approximately 40 to 46%. The judicial typology model, in contrast, was accurate in approximately 72 to 90% of cases tested (depending upon a variation in the model). The decision to reverse by a panel, and not just an individual judge, was also better explained by the judicial typology model. The presence of certain judge types on three-judge panels was associated

was but one factor that judges exhibited at different rates.

8 Corey Rayburn Yung, Flexing Judicial Muscle: An Empirical Study of Judicial

Activism in the Federal Courts, 105 NW. U. L. REV. (forthcoming 2011) [hereinafter “Yung Activism”]; Corey Rayburn Yung, Judged by the Company You Keep: An Empirical Study of the Ideologies of Federal Judges, 51 B.C. L. REV. 1133 (2010) [hereinafter “Yung Ideology”]; Corey Rayburn Yung, Playing Favorites and Reaching Consensus: An Empirical Study of Judicial Partisanship and Independence in the Federal Courts (contact Author for manuscript) [hereinafter “Yung Partisanship”].

9 Fischman & Law, supra note 4, at 168-170.


with decisions that respected deferential standards of review 15% more often that panels with other judge types. Further, evidence supported the typology-based prediction that panels with certain partisan judge types would apply standards of review in a manner that favored district judges appointed by one political party. In contrast, the ideology-based model performed in a manner wholly inconsistent with its expectations. The ideological makeup of the panel against the district court was uncorrelated with the decision by a panel to reverse.

The study further examined whether the judicial types identified were correlated with background characteristics of the judges studied. Notably, a judge’s type was correlated with, among other variables, ranking of law school attended, whether a judge had taken senior status, and the appointing President’s political party. By aggregating those three correlated variables, a predictive model was constructed and tested which forecasted how judges would behave on the bench based upon those background factors in combination.

Section I of this Article briefly discusses existing research and the dominant views pertaining to judicial decisionmaking, particularly for judges on the Courts of Appeals. In Section II, the underlying study design, data, and four behavioral measures (activism, ideology, independence, and partisanship) are described in detail. Section III outlines the statistical process used to identify the differing judicial styles and discusses the salient traits of each type that are supported by the data. In Section IV, the results of the tests of the nine judicial types to determine if they predicted judicial dissents and reversals as well as any connections between the categories and background traits of the judges are described in details. Section V details the limitations on the study and discusses validity and reliability of the methods and data used herein. The article briefly concludes with thoughts about the new directions that should be taken in empirical research on the basis of the findings in this study. Consistent with the mission of making empirical legal studies about law and courts more accessible to a broader audience,10

10 Lee Epstein, Andrew D. Martin, & Matthew M. Schneider, On the Effective

Communication of the Results of Empirical Studies, Part I, 59 VAND. L. REV. 1811, 1814 (2006) (“Most crucially, it seems nearly incontrovertible that moving towards more appropriate and accessible presentations of data will heighten the impact of empirical legal scholarship on its intended audience – be that audience other academics, students, policy makers, lawyers, or judges – not to mention raise the level of intellectual discourse among scholars themselves.”); Fischman & Law, supra note

this Article avoids empirical research jargon whenever possible

4, at 135-36 (“As a research community, we must cultivate and convey a better understanding of methods for measuring judicial ideology if we are to succeed in convincing others of the validity of our work.”).


and utilizes graphical representation11

of key measures throughout the Article. However, the statistical information traditionally found in empirical legal studies is located in the footnotes and appendixes at the end of this Article.

I. UNDERSTANDING JUDICIAL DECISIONMAKING Judges have been objects of study for decades, but there has been an

explosion in recent years of empirical research about how judges decide cases.12 In political science and law, there has been a particular focus on the role of political ideology in shaping judicial decisions.13 The Supreme Court has been the primary judicial body studied,14 often to the neglect of courts which have had a greater net impact on the development of American law.15

11 Lee Epstein, Andrew D. Martin, & Christina L. Boyd, On Effective Communication

of the Results of Empirical Studies, Part II, 60 VAND. L. REV. 801, 804-05 (2007) (“… researcher[s] should almost always graph their data and results…. Unless the author has a very compelling reason to provide precise numbers to readers, a well-designed graph is a superior choice to a table.”).

As a result, there has been a tendency by scholars to assume that the theories and models of Supreme Court decisionmaking should also be used

12 Gregory C. Sisk, The Quantitative Moment and the Qualitative Opportunity: Legal Studies of Judicial Decisionmaking, 93 CORNELL L. REV. 873, 874 (2008) (“In the past decade, the pace of empirical legal study has quickened, and the publication of empirical studies in law journals has increased.”); Jeffrey M. Chemerinsky & Jonathan L. Williams, Measuring Judges and Justice, 58 DUKE L.J. 1173, 1174 (2009) (“Empirical scholars have begun to train these same tools on the judiciary. They have studied topics ranging from the economic effects of judicial systems to the influence of ideology on judicial decisionmaking.”); Jack Knight, Are Empiricists Asking the Right Questions about Judicial Decisionmaking?, 58 DUKE L.J. 1532, 1535 (2009) (“Social scientists who study the courts employ an impressive array of statistical and mathematical approaches. This array has grown in variety and sophistication in the last decade.”).

13 Gregory C. Sisk & Michael Heise, Judges and Ideology: Public and Academic Debates about Statistical Measures, 99 NW. U. L. REV. 743, 744 (2005) (“… the seclusion of the ivory tower has been breached, as public attention has become increasingly focused upon studies that suggest the influence of ideological and partisan variables on the outcome of court cases.”).

14 Frederick Schauer, Incentives, Reputation, and the Inglorious Determinants of Judicial Behavior, 68 U. CIN. L. REV. 615, 621 (“Although there is not an enormous amount of empirical research on judicial behavior generally, traditionally the overwhelming bulk of what there is has been about the Supreme Court of the United States.”).

15 Richard A. Posner, Judicial Behavior and Performance: An Economic Approach, 32 FLA. ST. U. L. REV. 1259, 1273 (2005) (“[T]he Supreme Court reviews only a minute percentage … of court of appeals decisions. Entire fields of law are left mainly to the courts of appeals to shape.”).


to describe the behavior of judges in general.16 However, the modern Supreme Court has been highly atypical in many ways and studies have shown that what has held true for the Justices was largely inapplicable to lower courts like the Courts of Appeals.17 Nonetheless, the major decisionmaking models used to describe lower court behavior originated from and were largely shaped by empirical research of the Supreme Court.18

A. Law, Strategy, and Politics Three basic models have defined the debate regarding judicial

decisionmaking among legal and political science scholars. The attitudinal model has contended that judicial choices are best explained by ideology.19 The strategic model has identified institutional and personal incentives as the explanation of judicial behavior.20 The formal model has viewed doctrine, precedent, and interpretation of codified law as central to decisions by judges.21

The dominance of the attitudinal, strategic, and formal models of decisionmaking can best be understood through a brief history of their origins. Much of the modern empirical research about judges was initially derived from political science models of legislatures.

22

16 Tracey E. George, Developing a Positive Theory of Decisionmaking on U.S. Courts

of Appeals, 58 OHIO ST. L.J. 1635, 1665 (1998) (“The attitudinal and strategic theories of judicial behavior described above have been developed almost entirely through a consideration of the behavior of U.S. Supreme Court justices.”).

Such conceptions of decisionmaking relied on spatial models which sought to identify

17 Richard A. Posner, The Role of the Judge in the Twenty-First Century, 86 B.U.L. REV. 1049, 1054 (2006) (“The Supreme Court is of course not a typical American court. The federal courts of appeals and state courts have a more diverse and less political docket and are constrained by threat of reversal (though not state supreme courts when deciding state-law cases).”) [hereinafter “Posner 21st”]; Schauer, supra note 14, at 621 (“…whatever may be true of the Supreme Court regarding the primacy of sincerely held policy preferences over self-interest, it would be difficult to generalize to other courts.”);

18 George, supra note 16, at 1665. 19 See generally, JEFFREY A. SEGAL & HAROLD J. SPAETH, THE SUPREME COURT AND

THE ATTITUDINAL MODEL REVISITED (2002). 20 Posner 21st, supra note 17, at 1056. 21 See generally, Harry T. Edwards & Michael A. Livermore, Pitfalls of Empirical

Studies that Attempt to Understand the Factors Affecting Appellate Decisionmaking, 58 DUKE L.J. 1895 (2009); Harry T. Edwards, The Effects of Collegiality on Judicial Decision Making, 151 U. PA. L. REV. 1639, 1656 (2003); Harry T. Edwards, Collegiality and Decision Making on the D.C. Circuit, 84 VA. L. REV. 1335, 1336 (1998); Harry T. Edwards, Public Misperceptions Concerning the “Politics” of Judging: Dispelling Some Myths About the D.C. Circuit, 56 U. COLO. L. REV. 619, 625 (1985).

22 Knight, supra note 12, at 1536 (“Formal theorists who study the courts initially borrowed their analytical models from the research on legislative bodies.”).


ideological distance among individual legislators.23 The early studies of the courts viewed judicial decisions as policy choices similar to those made by legislators.24 The legislative roots in the modern judicial decisionmaking research helped spawn the attitudinal model.25

The attitudinal model was contrasted with the formal model of law which was perhaps more strawman than actual position taken by scholars.

26 Modern adherents to the formal theory of decisionmaking have not argued that every case has been decided solely by mechanical application of the law, but have contended that the large majority of legal disputes have been resolved by determinate rules utilized by neutral judges.27 The formalists have quite rightly cited the differences between the Supreme Court and other courts as to the centrality of ideology in decisionmaking.28

The strategic model addressed the deficiency concerning personal and institutional goals of judges that was missing from both the formal and attitudinal models. As Judge Posner has explained:

Yet it doesn't follow that judicial behavior is not determined by incentives. Without incentives, there is paralysis…. Judges have a utility function, as economists refer to a person's system of preferences, just like everybody else…. Clearly one that remains is leisure, and in the age of the law clerk the opportunities for a leisured judicial career are abundant. Yet most judges work pretty hard, and many work very hard indeed…. What are they working hard for? Some for celebrity, but most are content to labor in obscurity. I think most judges (I have in mind particularly federal appellate judges, the slice of the judiciary that I know best) are guided in their judicial performance primarily by two objectives that are different from and more interesting than a desire for leisure or a thirst for celebrity. One is a desire to change the world for the better (which to the cynical is simply a desire to exercise power - and the ability to exercise even modest power is indeed a perk of being a judge). The other is to play the judicial game.29

23 Id. (“The models were standard spatial models that sought to analyze legislative

votes as a function of the ideological distance between a legislator's ideal preference point and the utility point of the proposed piece of legislation under consideration.”); WALTER F. MURPHY & JOSEPH TANENHAUS, THE STUDY OF PUBLIC LAW 12 (1972).

24 Knight, supra note 12, at 1537 (“In these early studies judicial choices were operationalized, not as dispositional votes, but rather as measures of the substantive policy consequences of the decisions.”).

25 GLENDON SCHUBERT, THE JUDICIAL MIND: THE ATTITUDES AND IDEOLOGIES OF SUPREME COURT JUSTICES, 1946-1963 6 (1965).

26 H. Jefferson Powell, A Response to Professor Knight, Are Empiricists Asking the Right Questions about Judicial Decisionmaking?, 58 DUKE L.J. 1725, 1725 (“... the empiricists frequently appear to be battling a formalist straw man who believes that law can be done by following rules that do not allow for discretion in their interpretation or application.”).

27 Edwards, supra note 21, at 1897. 28 Id. 29 Posner 21st, supra note 17, at 1056.


Generally, advocates of the strategic model have not embraced a strong version of the theory that incentives have dictated all judicial decisions. Instead, like those who have supported the attitudinal and formal models, concessions have been made to incorporate the other theories.30

With the three global theories of decisionmaking in place, much of recent scholarship in the area has focused on determining the degree to which each one has explained greater portions of decisionmaking in different court settings.

31 Regarding the Courts of Appeals, Frank Cross gave the most comprehensive treatment of the three theories and found some evidence to support the attitudinal model, more to indicate formal constraints on decisionmaking, and almost no corroboration for the strategic theory.32

However, it also illustrated how data and model limitations have fundamentally constrained understanding of decisionmaking in the Courts of Appeals. Cross, like most every other researcher engaged in comprehensive study of those courts, used the United States Courts of Appeals Database (“Songer Database”), which included more than 18,000 opinions from 1925 to 1996.

Because Cross’s study was the first to systematically assess the three models as applied to the Courts of Appeals, its value in this area has been substantial. In some ways, it was the pinnacle of what could be done with existing data and the three major models in studying federal appellate judges.

33 Because the database covered the full range of circuits during such a long time frame, there were just a handful of opinions from each circuit in a given year. As a result, it was virtually impossible to use the Songer Database for individualized assessments of judges in any given time frame. Further, although the coding was very extensive in some areas (such as case issues), it was not able to be utilized at all for the ideology, activism, and partisanship metrics described in later this Article.34

30 Id. at 1054-56.

While the Songer Database has been an extremely valuable

31 See, e.g., George, supra note 16 (testing a combination of the attitudinal and strategic models in Fourth Circuit en banc cases).

32 Cross Decisionmaking, supra note 6, at 1514 (“The results of this study shed considerable light on the nature of judicial decisionmaking. The traditional legal model clearly explains a significant part of this decisionmaking, even after controlling for ideology and other variables. The legal model obviously leaves room for other considerations, though, since judicial ideology is also consistently a significant determinant of some decisions. The strategic model appears to explain little, if any, circuit court decisionmaking.”).

33 KUERSTEN & SONGER, supra note 7, at 1. 34 It might have allowed for the Independence measure, but the difficulties created by

the addition of the passing of time and small sample sizes for judges with limited times on the bench would problematize any results. Although the Songer Database included a field


resource, research based upon it has epitomized the homogeneity, unidimensionality, and separateness assumptions of judicial decisionmaking.

More recently, there has been an emergence of perspectives from other fields that have pushed the discussion beyond the three models which have reigned supreme in scholarly debates.35 Particularly notable has been Judge Richard Posner’s attempt to give readers a peak into judicial chambers in How Judges Think which identified nine theories of decisionmaking from a range of academic disciplines.36 In addition to the three major models, Posner discussed sociological, organizational, psychological, economic, pragmatic, and phenomenological theories.37 However, the additional theories investigated by Posner have largely not been subject to the rigor of empirical analysis at the level applied to the political science models.38

B. Testing Our Assumptions While the three models described above have predominated empirical

research by political science and legal scholars, there has also been a substantial body of work completed over the last thirty years which has recognized different types of judges. In 1981, J. Woodford Howard published a study based upon surveys of judges sitting on three federal Courts of Appeals.39 Howard identified three types of judges: 1) innovators who sought to inject new ideas into the law; 2) interpreters who believed that the role of the judge was limited to discerning meaning; and 3) realists that embraced portions of the other two theories.40 More recently, David Klein’s survey of federal appellate judges did not construct a formal typology, but acknowledged the diversity of approaches used.41

for standard of review, which was essential to the measures in this article, coding of it was limited to instances of agency review and not lower court review.

Jeffrey Berger and Tracey George found a particular type of judges, judicial

35 See, e.g., Dan Simon, A Psychological Model of Judicial Decisionmaking, 30 RUTGERS L.J. 1 (1998).

36 RICHARD A. POSNER, HOW JUDGES THINK (2008). 37 Id. at 19. 38 Id. at 19-56. 39 J. WOODFORD HOWARD, JR., COURTS OF APPEALS IN THE FEDERAL JUDICIAL SYSTEM

(1981). 40 Id. at 160-62. 41 DAVID E. KLEIN, MAKING LAW IN THE UNITED STATES COURTS OF APPEALS 21-23

(2002).


entrepreneurs, at the federal appellate level by utilizing citations of judges by the Supreme Court.42

Frank Cross and Stefanie Lindquist recently summarized the prevailing evidence concerning diverse judging styles that has not been interwoven into most empirical research about how judges render decisions:

Judges are a diverse lot, like any group of humans. They value different things in the course of rendering their decisions. Some may be more ideological in their judgments. Some may be more traditionally legalistic in their decisionmaking. Others may be pragmatic. As the surveys show, different judges view their roles differently. Judges place different emphases on the tradeoff between their roles as “generators of precedents” and resolvers of disputes.43

It is not at all revolutionary to recognize that judges from varied backgrounds, living in different geographic locations, nominated against distinct political conditions, and having a range of ideological and legal views, have different decisionmaking methodologies. However, empirical research has done little to build upon the recognized differences among individual judges (outside of Supreme Court studies).

Similar to work that questions the homogeneity assumption, efforts have been made to test the unidimensionality premise when possible.44 Further, panel effects research has also been extremely valuable in breaking down the idea that judges have been separate islands rendering independent decisions.45 However, such research has yet to incorporate multiple levels of the judiciary and only has examined the interrelations between judges on small panels.46

42 Jeffrey A. Berger & Tracey E. George, Judicial Entrepreneurs on the U.S. Courts of

Appeals: A Citation Analysis of Judicial Influence, available at http://ssrn.com/abstract=789544.

The study described in this Article began as an effort to expand the work of scholars that have tried to create measures and models that incorporated judicial heterogeneity, multidimensionality, and interconnectedness. In doing so, four separate measures of federal appellate judge behavior were created that rendered individual judge scores based upon interactions with co-panelists and district court judges.

43 Frank B. Cross & Stefanie Lindquist, Judging the Judges, 58 DUKE L.J. 1383, 1416 (2009).

44 See, e.g., Paul H. Edelman and Jim Chen, The Rehnquist Court in Empirical and Statistical Retrospective: The Most Dangerous Justice Rides into the Sunset, 24 CONST. COMMENTARY 199, 200 (2007) (“To account for and accommodate these alternative dimensions, our earlier research constructed a method for identifying the most powerful Justice without relying on the assumption of unidimensional policy preferences. Instead, earlier efforts focused on the unique policy coalitions formed by the Justices in non-unanimous cases.”).

45 See, e.g., Cross & Tiller, supra note 5, at 2168-75; Revesz, supra note 5, at 1751-56; see generally, Kim, supra note 5.

46 Id.


II. BEHAVIORAL MEASURES

The primary method of measuring behavior for the federal Courts of

Appeals has been static proxy techniques. For example, studies of judicial ideology have most commonly used the appointing President’s political party as a proxy for the judge’s ideology. There has been a growing call by some scholars to move from such static proxy measures of judges to metrics derived from the actual behavior of judges.47 Among other advantages, behavioral measures allow scholars to comparatively test model performance instead of being merely satisfied with finding statistical significance between proxy metrics and outcomes.48

As the results of this study indicate, there is strong reason to believe that measuring judicial performance can be a much better predictor of future behavior than traditional proxy metrics.

A. Data Gathering and Database Construction The underlying data used in this study has been used by the Author in

related research,49 but several important changes have been made that are noted below. Data were gathered from published and unpublished 2008 opinions issued by the First, Second, Third, Fourth, Fifth, Sixth, Seventh, Eighth, Ninth, Tenth, and Eleventh Circuit Courts of Appeals. The analyzed dataset (“Case Database”) from those circuits included over 10,000 panel decisions and, as a result, over 30,000 votes by judges sitting on the Courts of Appeals.50 The Case Database was created from searches of LexisNexis databases of the Courts of Appeals that included standard of review language,51 and excluded immigration52 and habeas corpus53

47 Fischman & Law, supra note

cases,

4, at 170. 48 For a general critique of the dangers associated with overreliance on statistical

significance as the focal point of research, see generally STEPHEN T. ZILIAK & DEIRDRE N. MCCLOSKEY, THE CULT OF STATISTICAL SIGNIFICANCE: HOW THE STANDARD ERROR COST US JOBS, JUSTICE AND LIVES (2008).

49 See Yung Activism, supra note 8; Yung Ideology, supra note 8. 50 The normal assumption that there would be three judicial votes for each panel in the

dataset slightly overestimates the number of actual votes. There were a few instances where only two judges issued an opinion due to a death or recusal of a panel member. There were also instances when a single judge would issue an order without sitting on a three-judge panel.

51 For each of the Courts of Appeals databases in LexisNexis, the following search was executed and all of the results were downloaded and coded: “date aft 1/1/2008 and date bef 1/1/2009 and (“De Novo” or Clear! Erro! or (Arbitrar! w/3 Capricious!) or (Abus! w/3 Discretion) or “Substantial Evidence” or “Standard of Review"”)) and not immigration and not habeas.”


which contained unique standard of review issues. Among other variables, cases were coded for: judges on the panel, whether individual judges were sitting by designation, appellate disposition, appellant, appellee, type of case (e.g., criminal or environmental), prevailing party, circuit, district court judge, district court, whether the case involved review of an executive agency decision, and standard of review used. In analyzing each case, the vote of each judge on the panel was coded separately, and thus a dissent by a judge in a case was coded so that the dissenting judge had the opposite view and disposition as the panel majority. This allowed for each judge’s behavior to be studied independently and collectively. The more recent version of the dataset was re-coded using off the shelf computer software54 and scripting in Visual Basic.55

52 See Hiroshi Motomura, Immigration Law and Federal Court Jurisdiction Through

the Lens of Habeas Corpus, 91 CORNELL L. REV. 459, 474 (2006) (discussing how large portions of immigration case review—those by the Board of Immigration Appeals—are based upon a collateral review model, which affords a very high level of deference because “[t]he regulations . . . revise the standard of review to require greater deference to an immigration judge's findings of fact.”).

The automated coding process searched for over 100 relevant variables in the header fields and text of opinions. Human coding was used to supplement and quality check the automated coding.

53 See Brandon Scott, When Child Abuse Becomes Child Homicide: The Case of Gilson v. Sirmons, 34 OKLA. CITY U. L. REV. 281, 293-94, 305 (2009) (discussing the “unique” standard of review in federal habeas cases due to the Antiterrorism and Effective Death Penalty Act).

54 TextConverter 3, available at: http://simx.com/. 55 Corey Rayburn Yung, Automated Coding of Federal Court Cases (contact author for

manuscript). In the prior studies, the coding of the data was performed exclusively by law student research assistants and the Author. Due to inconsistencies among and within the Courts of Appeals in formatting opinions, the automated coding was not foolproof. However, quality checking has indicated that the automated coding for every variable used in this study was superior to the human coding done in prior editions of the dataset.

Because the prior human coding had already been performed for over 96% of the cases in the new dataset, the human and computer coding were integrated. In particular, because some circuits have not regularly or ever reported the identity of district court judge whose judgment was reviewed, human coding was essential in that area. When there were discrepancies in coding between the two systems, the differences were resolved by the Author. The automated coding also allowed for coding of many new fields, primarily in the area of case issues. This provided more detailed information that could be used in adjusting various behavioral scores based upon differing case issue mixes of judges and circuits. Other changes in the coding, which may have accounted for some of the different scores used in this study included: separation of decisions to vacate, reverse, remand, and reverse in part, coding of concurrences, to the degree possible determining which standard of review was actually applied when multiple standards were listed, corrections of errors in the source data downloaded from LexisNexis, further quality checking, coding of the actual opinion writer (as opposed to merely including them in the majority), and removal of votes recorded on the decision to not hear a case en banc.


In addition to the Case Database, a separate database (“Judge Database”) was constructed which included biographical and other data about individual judges. In the Judge Database, judges were coded for, among other variables: appointing president, presidential party, age at the time of appointment, age in 2008, composition of the Senate at the time of appointment, gender, race, law school attended, prior work experience, whether the president and majority of the Senate were of the same party at the time of appointment, and whether the judge took senior status during or before 2008. The Judge Database included data for all federal appellate judges that served on panels as well as district judges who had opinions reviewed in the Case Database. In all, background trait information was accumulated for over 1,500 federal judges. Data in the Judge Database was entirely human coded from publicly available information about the judges studied.

B. Decisionmaking Metrics

Because the decisionmaking measures discussed below have been

expounded upon in prior scholarship, the discussions herein are limited to the necessary elements of each and the basic methodologies used for their computation. In some instances, notes about the modifications to previously methodologies have been included as well. The four dimensions isolated and measured in this new research were activism, ideology, independence, and partisanship.

1. Activism and Deference

One of the most commonly used words to describe judicial behavior in

public discourse is “activism.”56 The trait associated with activism has been exhibited when a judge placed his or her own judgment above others.57 It has been indicated by a preference against judicial deference wherein a judge, in the aggregate, placed his or her views above legislatures, people in the executive branches, and other judges.58 When viewed in this way, “activism” could be thought of a concept that was morally-neutral.59

56 Keenan D. Kmiec, The Origin and Current Meanings of “Judicial Activism,” 92

CALIF. L. REV. 1441, 1443-44 (2004).

The

57 Yung Activism, supra note 8, at __. 58 Id. 59 Ernest A. Young, Judicial Activism and Conservative Politics, 73 U. COLO. L. REV.

1139, 1144 (2002) (“[J]udicial ‘activism’ is not just an epithet—or at least . . . it does not have to be. Properly used, ‘activism’ and ‘restraint’ focus us on the institutional aspects of judicial decisions, apart from the merits, and this focus will be useful from time to time.”)Robert Justin Lipkin, We Are All Judicial Activists Now, 77 U. CIN. L. REV. 181,


measurement of activism described below was merely a relative assessment of how likely a judge was to defer compared to other judges studied.

In order to be properly measured, a few minor refinements to the general definition of “activism” were made to provide a baseline for comparison. Otherwise, the ability to separate “activism” from “deference” would have been lost. Ultimately, the definition used for empirical study was as follows:

Activism occurs when judges do not defer to constitutionally significant actors when a formal model of legal decisionmaking would predict otherwise.60

While it has been generally difficult to reach anything approaching consensus on what a formal model of law would predict, there has been a useful analytical tool of the Courts of Appeals that provides an effective baseline.61 Standards of review have been used by federal appellate courts to indicate when they should defer to lower court judgments and when no deference is required.62

A judge who has regularly discounted deferential standards of review by reversing lower court judgments can be said to be more activist in the aggregate than a judge who has reversed less often in such situations. However, to remove ideological differences and other strategic motivations from the concept of “activism,” it was necessary to compare a judge’s behavior when using deferential and non-deferential standards of review. When applying a de novo standard, judges were expected to reveal their true preferences. In contrast, it was expected that those preferences would be suspended from time to time when applying a deferential standard. Based upon those premises, the basic measurement for activism used in this study was:

Raw Activism Differential = Reversal rate using deferential standard – Reversal rate using de novo standards63

203 (2008) (“Nothing in this distinction suggests that either form of judicial activism is good or bad. It is a distinction that should be used as a filter through which to evaluate different proposals for defining judicial activism.” (footnote omitted)).

60 Yung Activism, supra note 8, at __. 61 For a discussion of the general neglect of baselines in empirical legal research and

the problems it has created, see generally Brian Z. Tamanaha, Devising Rule of Law Baselines: The Next Step in Quantitative Studies of Judging, THE LEGAL WORKSHOP (Mar. 25, 2010), http://legalworkshop.org/2010/03/25/2667 (submitted by DUKE L.J.).

62 Frank B. Cross, Decisionmaking in the U.S. Circuit Courts of Appeals, 91 CALIF. L. REV. 1457, 1502-03 (2003).

63 Yung Activism, supra note 8, at __.


Thus, a judge with a higher Activism Differential exhibited greater activism because he or she was reversing lower court decisions using a deferential standard more often relative to reversing in unconstrained situations. The Raw Activism Differential was adjusted to reflect the particular circuit on which a judge sat, the mix of case issues each judge had, any “super-panel” effects64 that might have been experienced by the judge, and scaling to a common system. Greater detail about these adjustments can be found in Appendix A.65

2. Ideology – Liberal to Conservative Judicial ideology has been the primary target of those seeking to

measure judicial behavior.66 Even when it has not been the focus of particular studies, it has almost always been essential to control for it.67 Although there has been some variation in definitions of ideology,68

there has been substantially less controversy about the concept than there has been concerning activism. The very basic definition used in this study was:

Judicial ideology is the set of political beliefs that guide judicial decisions in some cases.69

This study utilized what has become known as an agnostic measure.70

64 Id.

Such measures have relied on the idea that those who have similar ideologies would vote together more often and those who have dissimilar ideologies will disagree more often. If enough data was gathered, a researcher could identify groups with similar ideologies. With some prior assessments about who is more likely to be conservative and liberal, agnostic scoring systems could be applied. Agnostic measures have had the advantage that they could be applied to large populations of data far more easily than other behavioral techniques. A researcher could merely identify the disposition of an appellate panel’s judgment (e.g., affirmed, reversed, etc.) and any voting blocs among the panelists. However, at the Courts of Appeals, almost 98% of panels in recent years have voted unanimously which has prevented the use of agnostic measures entirely for those courts.

65 The mean Activism Score was 51.3 and the standard deviation was 16.7. The data was scaled linearly where a 0 score corresponded with a -32.4% adjusted Raw Activism Differential and a 100 score indicated a +23.2% adjusted Raw Activism Differential.

66 Fischman & Law, supra note 4, at 135. 67 Id. 68 Id. 69 Yung Ideology, supra note 8, at 1140-44. 70 Fischman & Law, supra note 4, at 135.


By including the district court judge in the “super-panel” of judges, greater information was gained and the unanimity problem resolved. Unlike dissents on appellate panels, reversals of district court judgments have been far more common. As a result, integrating disagreements between appellate judges and district court judges gave the study more points of comparison for the creation of an accurate ideology scoring system. When a judge on a federal appellate court reviewed the judgment of district court judge, the appellate judge had three interactions which potentially yielded information about the appellate judge’s ideology (two interactions with co-panelists and one with the district judge). Using this agnostic coding system, the basic measurement of a judge’s ideology was as follows:

Raw Ideology Score = Rate of Agreement with Judges Appointed by Republican Presidents – Rate of Agreement with Judges Appointed by Democratic Presidents Numerous adjustments were made to the Raw Ideology Score to provide

a valid measure. Most significantly, because the district judge was situated in a different role than co-panelists, the rate of agreement with district judge was measured separately. As a result, agreements with the district judge in the form of affirmances were adjusted by standard of review used and type of case (criminal or civil). A Raw Ideology Score was then computed for both the panel and district judge interactions. Those scores were combined and further adjusted to include specific circuit adjustments, further case issue mix adjustments, any “super panel” effects experienced by the judge, and scaling to a common system. These later adjustments are described in greater detail in Appendix A.71

3. Independence and Collegiality Independence has recently been identified in recent studies as a

characteristic of judicial behavior worthy of substantial empirical investigation.72

71 The mean Ideology Score was 2.9 and the standard deviation was 20.9. The data was

linearly scaled where a -100 corresponded with a -21.4% adjusted Raw Ideology Score and a 100 indicated a 15.5% adjusted Raw Ideology Score.

This study conceived of and measured independence in a similar, but distinguishable manner, from prior attempts. Independence was not viewed as a positive or negative quality for a judge – it was simply

72 See, e.g., Stephen Choi & G. Mitu Gulati, Essay, A Tournament of Judges?, 92 CAL. L. REV. 299 (2004) [hereinafter “Choi & Culati Essay”]; Stephen J. Choi & G. Mitu Gulati, Choosing the Next Supreme Court Justice: An Empirical Ranking of Judge Performance, 78 S. CAL. L. REV. 23 (2004); Stephen J. Choi & G. Mitu Gulati, Mr. Justice Posner? Unpacking the Statistics, 61 N.Y.U. ANN. SURV. AM. L. 19 (2005).


another trait that described tendencies in behavior. This basic definition of judicial independence was used:

A more independent judge is one that is more apt to provoke minor and major disagreements with other judges beyond what ideological differences would predict.73

Two types of disagreements were incorporated in the measurement applying the definition: concurrences and dissents. Unlike prior research, this study recognized that when other judges dissented from or concurred with a studied judge, that action was also in indicator of the studied judge’s independence. The following formula was used to determine the Raw Independence Score:

Independence Score = Rate of Dissents + Rate of Concurrences – Expected Rate of Dissents – Expected Rate of Concurrences

To remove disagreements based upon ideological differences, it was

necessary to determine the rate at which ideological variance accounted for concurrences and dissents so that the Independence Score could be properly untangled from the Ideology Score. The projected dissent and concurrence rates for each judge were computed and subtracted from the observed rates. Three additional adjustments were made to the Raw Independence Scores. The values were adjusted so that each judge the scores were based upon an average criminal and civil case mixes, placed on a logarithmic scale (to create a normal distribution), and scaled to a common system.74 The case mix and common scaling adjustments are further explained in Appendix A.75

4. Partisanship and Indifference Partisanship has not been considered a separate trait from ideology and

activism in the rare instances that it has been referenced in empirical research. When it has been used to describe or study judicial behavior, it has

73 Yung Partisanship, supra note 8, at __. 74 Because of limited population sizes for judges traveling to other circuits when

concurrences and dissents were issued, a separate circuit adjustment was not performed for the Independence Scores.

75 The mean Independence Score was 44.1 and the standard deviation was 15.3. The data was logarithmically scaled, but before scaling a 0 corresponded with a -10.0% adjusted Raw Independence Score and a 100 indicated a 24.9% adjusted Raw Independence Score.


normally been a negative attribute indicating ideological decisionmaking.76

As with the other measures, the goal of this study was to treat partisanship (and its opposite, indifference) as morally neutral behavior exhibited by judges on a relative scale. However, like the word “activism,” “partisanship” has carried significant negative connotations. Despite that shortcoming, “partisanship” and “indifference” were chosen because they best reflected the behavior being studied. The basic definition of partisan behavior used in the study was:

A judge is acting in a more partisan manner when he or she applies a neutral legal rule in a manner that demonstrates greater deference to members of a particular political party.77

Like the Activism Score, the Partisanship Score focused on the standard of review as the baseline for identifying partisan behavior. If a judge deferred more often to a district judge appointed by the same political party more often, he or she was more partisan than a judge who deferred at the same rate. And similar to the Activism Score, the focus of the partisanship measure was exclusively based upon appellate and district court interactions. The following formula provided the Raw Partisanship Score:

Partisanship Score = (Deferential Reversal Rate of Democratic District Judges – Deferential Reversal Rate of Republican District Judges) – (Non-Deferential Reversal Rate of Democratic District Judges – Non-Deferential Reversal Rate of Republican District Judges)78

Several adjustments were made to the Raw Partisanship Score. The one unique alteration was related to the inclusion of the appointing party of the Court of Appeals judge so that a Democratic appellate judge who favored Republican district judges (not partisan) would be scored differently than a Democratic appellate judge who favored Democratic district judges (partisan). The Raw Partisanship Score was adjusted to reflect the particular circuit on which a judge sat, the mix of case issues each judge had, any “super-panel” effects that might have been experienced by the judge, and then scaled to a common system. Greater detail about these adjustments can be found in Appendix A.

Notably, the study identified judges of all combinations of high and low Partisanship, Activism, and Ideology Scores. For example, Justice, then-

76 See, e.g., Daniel E. Ho & Kevin M. Quinn, How Not to Lie with Judicial Votes:

Misconceptions, Measurement, and Models, 98 CALIF. L. REV. 813, 817 (2010). 77 Yung Partisanship, supra note 8, at __. 78 Id., at __.


Judge, Sonia Sotomayor had a moderate Ideology Score (17.3), a low Activism Score (39.3), and a Partisanship Score nearly one standard deviation above the mean (57.2). Further, based upon robust linear regression, the Activism, Partisanship, and Ideology Scores were not correlated in a statistically significant manner.79 So, the data indicated that the Partisanship Score was describing some separate behavior from the Ideology and Activism Scores.80

III. JUDICIAL STYLE TYPOLOGY “Style” might seem an odd word to apply to what judges do because it

lacks the grandiosity of “philosophy” and seriousness of “methodology.” However, because this study was focused on measuring the behavior of judges as manifested in outputs (judicial opinions and votes), it was the most appropriate label. Judges might have had elaborate decisionmaking methods and/or committed judicial philosophies, but that does not mean such perspectives have been consistently exhibited in opinions and votes. This could have occurred for several reasons. Judges may not have practiced what they preached. Their decisionmaking technique might have only been applicable in some cases. The methods or philosophy might have pointed in conflicting directions. Further, there have been just a handful of judges with public records about their philosophies or methodologies from which any information could reliably and validly inferred.

In contrast, “style” focused exclusively on the outward appearances of the judicial decisionmaking. No references to or inferences about the decisionmaking process were made by the label “style.” Instead, the label stayed true to the study methodology which only considered the end result of the decisionmaking process and did not in any way attempt to determine how such decisions were reached beyond the opinions published.

A. Multidimensional Clusters of Judges

To construct a viable typology of judges, the 178 studied judges needed

to be sorted into groups based upon common traits. The technique by which this was done was cluster analysis. Cluster analysis is a data analysis tool

79 For Activism and Ideology Scores, p = 0.6530; Activism and Partisanship Scores, p

= .1754; Ideology and Partisanship Scores, p = 0.4057. Robust regressions were also done to compare the Ideological Scores (absolute value of the Ideology Scores) with the Partisanship Scores. For Activism and Ideological Scores, p = 0.7443; Partisanship and Ideological Scores, p = 0.8418.

80 The mean Partisanship Score was 44.8 and the standard deviation was 15.3. The data was linearly scaled and a 0 corresponded with -73.7% adjusted Raw Partisanship Score and a 100 indicated an 89.8% adjusted Raw Partisanship Score.


which categorizes data from one variable based upon similarities and differences across a set of other variables.81

The particular cluster analysis technique used was Ward cluster analysis.

In this case, judges were grouped according to their Activism, Ideological, Independence, and Partisanship Scores. The Ideological Scores were simply the absolute values of the Ideology Scores such that a judge with a -60 score would be treated the same as one with a 60 score. This was done to find commonalities in judicial style among persons on opposite ends of the ideological spectrum. All four Scores were on a scale of 0 to 100, with 100 indicating the highest level of that trait within the 178 judges.

82 The Ward technique focused on minimizing variance when placing a judge in one group as opposed to another.83

Ultimately, based upon careful review of the cluster tree that was created, nine distinctive clusters of judges were identified. These groups were then labeled according to their mean average scores. The nine judicial types were: Trailblazing, Consensus Building, Stalwart, Regulating, Steadfast, Collegial, Pragmatic, Minimalistic, and Error Correcting. Individual judges in each group might have had scores which deviated by a reasonable margin from the category mean, but the Ward technique placed them in the appropriate group to minimize the variation versus placing the judge in an alternate group. Notably, as illustrated in Figure 1 below, the distribution among the judge style types was not even.

Because the clusters were formed along four separate dimensions, it was impossible to include a visual depiction of the shapes that enveloped the various clusters in a two dimensional medium such as this. However, the commonalities among judges within each group can be seen in Appendix B where all of the judge scores are listed.

81 See generally, BRIAN S. EVERITT, SABINE LANDAU & MORVEN LEESE, CLUSTER

ANALYSIS (4th ed. 2001); Frank B. Cross & Stefanie Lindquist, Judging the Judges, 58 DUKE L.J. 1383, 1425 (2009) (“Cluster analysis is a statistical tool that enables like items to be grouped based on similarities and dissimilarities across a set of variables…. The procedure identifies the cases (individual judges) who share similarities with respect to certain variables ... and groups them into categories, creating a sort of taxonomy of judges.”).

82 For a prior example of the use of Ward cluster analysis in a judicial decisionmaking study, see Cross & Lindquist, supra note 43, at 1425-30.

83 CHARLES ROMESBURG, CLUSTER ANALYSIS FOR RESEARCHERS 129-35 (2004)


Each of the nine judging styles and the quantitative aspects that define each type are discussed in more detail below.

B. Judicial Styles

Among the nine judicial types, there were some groups which have

greater similarities than others. When possible, efforts were made to explain the critical differences among relatively similar groups. Further, the order of the styles discussed below was the one created by the statistical cluster analysis. Thus, to a degree the order reflects how the groups were created with consecutive categories often, but not always, having commonalities.

42

34 34 32

16

7 5 5 3

05

1015202530354045

Num

ber o

f Jud

ges

Figure 1 - Distribution of Judges among Style Categories


1. Trailblazing

Trailblazers exhibited higher Ideological Scores (usually over two

standard deviations from zero), higher Independence Scores, lower Activism Scores, and lower Partisanship Scores. They had the second highest Ideological Scores of any group, but, in contrast to conventional wisdom, their behavior was less partisan and activist than the average judges. It might be thought that these judges were primarily in the business of carving out ideological decisions that shaped circuit law and invited opposition by co-panelists. Yet, the review of district courts was less significant to Trailblazers.

53.4

33.1 37.2

61.3

14.0

51.344.8 44.1

0

20

40

60

80

100

Ideological Activism Partisanship Independence

Ave

rage

Sco

re

Figure 2 - Average Trailblazing Judge Scores

Trailblazing Average


2. Consensus Building

Consensus Builders, like Trailblazers, had very high Ideological Scores.

Yet, again contrasting with traditional perspectives on decisionmaking, they had the lowest Independence Scores (with typically below average Activism and Partisanship Scores). If conventional theories regarding the role of judicial politics were applied, it would have been expected that the most ideological judges would be more likely to dissent, concur, and invite concurrences and dissents. That this group maintained strong ideology while inviting the least relative disagreement among the judges indicated that they might be better at building voting blocs among ideologically diverse judges.

45.0 39.5 37.9

23.114.0

51.344.8 44.1

0

20

40

60

80

100


Ave

rage

Sco

reFigure 3 - Average Consensus Building Judge Scores

Consensus Building Average


3. Stalwart

Stalwart judges had the highest Ideological, Activism, and Partisanship

Scores of any judge type. Interestingly, though, their Independence Scores were usually below the mean. Stalwarts were heavily involved in the review of district court judgments in a manner indicating a partisan bent. Yet, despite the clearest ideological convictions of any judge type, they did not draw dissents and concurrences at rates expected by such strong ideological views. Stalwarts appeared to be the judges most consistent with the classic attitudinal model.

4. Regulating

71.7

88.3

71.5

33.6

14.0

51.344.8 44.1

0

20

40

60

80

100


Ave

rage

Sco

reFigure 4 - Average Stalwart Judge Scores

Stalwart Average

15.0

56.0 61.555.3

14.0

51.344.8 44.1

0

20

40

60

80

100


Ave

rage

Sco

re

Figure 5 - Average Regulating Judge Scores

Regulating Average


Regulating judges typically had Activism, Partisanship, and Independence scores above the mean and average ideological scores. The partisanship ratings were usually the highest for such judges. This meant that regulating judges were typically more likely to reverse the judgments of district court judges in situations where deference might be expected. And when such judges substituted their judgments in place of the district judge’s, the appellate judge typically did so in a partisan manner in the aggregate. It might be thought that these judges exercised their power primarily by keeping the district courts in line in a somewhat partisan manner.

5. Steadfast

Steadfast judges had the highest Independence Scores of any type,

elevated Activism Scores, and average Ideological and Partisanship Scores. Steadfast judges might be thought of as the relative lightning rods of the federal appellate courts because they were more likely to disagree with or invite disagreement by other judges (including district judges). Yet, there was no indication that such judges did so in a particularly ideological manner or that ideology was the source of disagreement with other judges. Their numerous disagreements apparently stemmed from other sources or were derived from limited willingness to compromise their points of view.

18.5

64.0

37.4

81.6

14.0

51.344.8 44.1

0

20

40

60

80

100


Ave

rage

Sco

re

Figure 6 - Average Steadfast Judge Scores

Steadfast Average


6. Collegial

Collegial judges, on average, exhibited scores below the mean in every

category. Judges in this category might be thought as adhering to a strong norm of consensus that has existed in the federal appellate courts.84

And their regulation of the district courts was relatively modest. Their overall profile might indicate a preference for less disagreement and conflict among federal judges.

84 See generally, Joshua B. Fischman, Decision-Making Under a Norm of Consensus:

A Structural Analysis of Three-Judge Panels, available at: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=912299.

6.8

49.237.8 38.9

14.0

51.344.8 44.1

0

20

40

60

80

100


Ave

rage

Sco

re

Figure 7 - Average Collegial Judge Scores

Collegial Average


7. Pragmatic

Pragmatic judges typically had elevated Partisanship scores and

Activism, Ideological, and Independence Scores that were slightly lower than or at the mean. Pragmatists were not true ideological or partisan moderates, but their overall profile indicating a relatively restrained stance in ideological disputes. Their lower Independence Scores indicated that they did not pursue conflict if it could be avoided. Unlike regulating judges, skeptical judges were not typically as partisan and were far less activist.

8. Minimalistic

12.0

44.853.1

32.8

14.0

51.344.8 44.1

0

20

40

60

80

100


Ave

rage

Sco

reFigure 8 - Average Pragmatic Judge Scores

Pragmatic Average

10.323.6

38.751.7

14.0

51.344.8 44.1

0

20

40

60

80

100


Ave

rage

Sco

re

Figure 9 - Average Minimalistic Judge Scores

Minimalistic Average


Judicial minimalism was the one style discussed herein that has received significant theoretical attention by scholars.85

In this study, such judges had the lowest Activism Scores, slightly higher Independence Scores, and close-to-average Ideological and Partisanship scores. The behavior of Minimalists should the strongest adherence to norms of deference in place of partisanship and ideology.

9. Error Correcting

Error correctors exhibited higher levels of activism than average,

slightly lower partisanship and ideology levels, and independence measures that were close to the mean. As a group, they were less likely to defer to district judge judgments, but generally more likely to apply deferential rules in a party-neutral manner. They were similar to regulating judges, but were significantly less partisan in reviewing district court judges.

IV. TESTING THE JUDGE TYPES

While the identification of distinctive judging styles might have been

worthwhile in its own right, it was also important to consider how different types of judges might have influenced decisionmaking. Testing the judging types in predicting judicial actions allowed it to be compared with the dominant explanations of decisionmaking. Further, it was investigated whether biographical or demographic characteristics of judges were correlated with their judging type. As a result, regression and other

85 See, e.g., CASS R. SUNSTEIN, ONE CASE AT A TIME: JUDICIAL MINIMALISM ON THE

SUPREME COURT (1999)

7.9

68.3

32.843.4

14.0

51.344.8 44.1

0

20

40

60

80

100


Ave

rage

Sco

re

Figure 10 - Average Error Correcting Judge Scores

Error Correcting Average


statistical analyses were performed to determine how judging types were connected to case outcomes and background factors.

A. Predicting Outcomes

The most effective way to determine the meaningfulness of the nine

identified judicial styles toward understanding judicial behavior was to test how well they predicted judicial behavior. Unfortunately, in order to do so required a dataset wholly separate from the one that was used to create the underlying measures of activism, ideology, independence, and partisanship. Otherwise, the test results would have been entirely circular because the input cases would have been the same as those used to judge the efficacy of the output values.86

As a result, a separate dataset was constructed for Courts of Appeals cases in 2009 (“Testing Database”). The basic coding process was similar to that for the 2008 case data except the 2009 data was solely computer coded.

As noted earlier, however, there was an inevitable difficulty in using another dataset because there simply was no publicly available data that had been coded to properly test the judicial styles.

87

In empirically evaluating the various judging styles, there were two major issues to address. First, it was important to determine if the judging styles were predictive of some individual decisions of judges on the Courts of Appeals in 2009. It was expected that most, if not all, would behave in manner, relative to the other judges, that was consistent with past behavior in 2008. Second, building upon the insights of prior research into so-called “panel effects,”

However, because circuits varied in their reporting of district court judge information, the ability to test certain aspects of the judicial types was limited by the available data. After removal of en banc opinions and other extraneous and problematic cases, the Testing Database contained over 25,000 cases (and over 75,000 votes by judges sitting on the Courts of Appeals). For the tests described below, however, only portions of the Testing Database were used as warranted by the parameters of the tests.

88

86 Fischman & Law, supra note

it was important to consider the effects of a judicial style on co-panelists. For example, a consensus building judge might be expected to decrease the rate of dissent and concurrence on panels even with ideological divisions. However, this study went beyond the scope of prior research by including the district court judge being reviewed in considering the interconnectedness effects among the judges. A variety of hypotheses

4, at 179. 87 Human error checking, correcting of variations on judge names, and excluding of

certain cases was performed by the author after the automated coding was complete. 88 Cross & Tiller, supra note 5, at 2168-75; Revesz, supra note 5, at 1751-56; Kim,

supra note 5.


were constructed and tested to identify possible correlations between judging types and judicial behavior. Further, a competitive model relying on traditional measures of ideology was used for comparison in each test.

The two types of behavior that could be studied within the contours of the Testing Database were the decision of an individual judge to dissent and a panel of judges to reverse the judgment of the district court. Each of those situations presented different methodological obstacles, but tests were constructed to evaluate the competing models. The results for the dissent and reversal tests are described below.

1. Predicting Dissents

Certainly among the three appellate panelists, the most noteworthy

decision an individual judge could have made was to dissent. Thus, much empirical research has focused on the studying of and inferring from judicial dissents.89 However, other than theoretical ideal point models, little has been done to predict which judges would actually dissent on the federal Courts of Appeals. Overall, dissent has been a very infrequent event which has been difficult to study. In the 2009 Testing Database, dissents occurred in just over two percent of cases. It has been this low level of disagreement that has led some scholars to argue that there have been substantially more hidden dissents that were suppressed for strategic or norm-based reasons.90

Nonetheless, testing was performed to determine model consistency with dissent patterns in the Testing Database. Because of the difficulties in finding distinctions among model performance with the highly unlikely event of a dissent, the Testing Database was limited to situations where: 1) a judge dissented; 2) one of the panelists was appointed by a President whose party was different than the other two; and 3) a Steadfast, Trailblazing, and/or Regulating judge was on the panel. Those criteria yielded 253 total cases. In each situation, the judge type and ideology models were evaluated based upon whether their predictions of the dissenter were correct. The baseline for model effectiveness was simple chance. A random guess would be able to predict the identity of the dissenter in one-third of cases. The ideology model was determined to be correct if the

If a researcher merely wanted to predict which panels in a population would have a dissenting opinion, he or she could assert that there would never be a dissent and be correct almost 98% of the time. As a result, against such a baseline regression, analysis in this area has been prone to error.

89 Richard A. Posner, What Do Judges and Justices Maximize? (The Same Thing

Everybody Else Does), 3 S. CT. ECON. REV. 1, 20 (1993) 90 Joshua B. Fischman, Understanding Voting Behavior in Circuit Court Panels,

available at: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1636002.


dissenter was the judge who was of the minority appointing President’s party. The judge typology model was deemed correct if the dissenter was one of three chosen judging types (Steadfast, Trailblazing, and Regulating). The following hypotheses were tested:

Hypothesis: Steadfast, Trailblazing, and Regulating judges would be more likely to dissent than other judging types. Hypothesis: The judicial typology model would better predict the identity of dissenting judges than ideological proxy measures and random guesses. Figure 11 below illustrates the results of the ideology model, the judge

type model, and the chance model (randomly predicting the dissenter)91

:

The ideology based model only was a modest improvement over chance as it yielded just a seven percent gain in accuracy. However, the judicial typology model was over twice as effective as chance and accurate in 76.7% more cases than the ideology-based model. Limiting the tests to situations where only the two most independent judging types were used (Steadfast and Trailblazing) yielded an even more accurate model. Figure 12 below includes the results in those cases:

91 The chance model was not actually implemented by using a pseudo-random number

generator. Instead, it was presumed that the result would be for 1/3 of the guesses to be accurate.

71.9%

40.7%

33.3%

0% 20% 40% 60% 80% 100%

Typology

Ideological

Chance

Dissenter Prediction Rate Consistent with Model

Mod

el U

sed

Figure 11 - Model Consistency Rate for Predicting Dissenter on Panel


When Trailblazing and Steadfast judges were on a panel that had a dissent, they collectively accounted for 89.7% of those dissents. The results indicated that the typology model predicted the correct outcome in 94.2% more cases than the ideological proxy of party of the appointing President. 2. Predicting Reversals

Unlike dissents, reversals of district court and executive agency courts

occurred with relative normalcy in the Courts of Appeals. As a result, there was far more data available to be used in testing. Using the 2009 Testing Database, there were 5,643 cases where: 1) judging types were known for all three judging types;92 2) party of the appointing President was available for the panelists;93 and 3) party of the appointing President was known for the district court judge.94

92 The most common reason for why a judging type would not be known is that a

district court judge would be sitting by designation on the panel.

The following hypotheses were tested:

93 For any Courts of Appeals judge serving during this time, the Judge Database included the relevant information. However, in some instances of judges sitting by designation, such background information was not available. For example, in the Second Circuit, judges from the Court of International Trade would often sit by designation. Because judges on that court have a different appointment process, panels with Court of International Trade judges were omitted from this testing.

94 Because the 2009 Testing Database was created based upon automated coding, it was wholly dependent upon what was contained within the appellate judicial opinions. Notably, some circuits do not list the district court judge in the opinion (e.g., the Fifth and Tenth Circuits). The Eighth Circuit commonly omits the district court judge name when the lower court judgment is reversed, but includes it in most affirmations. As a result, the

89.7%

46.2%

33.3%

0% 20% 40% 60% 80% 100%

Typology

Ideological

Chance

Dissenter Prediction Rate Consistent with Model

Mod

el U

sed

Figure 12 - Model Consistency Rate for Predicting Dissenter on Panel


Hypothesis: Stalwart, Steadfast, and Error Correcting judges would defer less often to district court judgments using deferential standards of review than other judge types. Hypothesis: The judicial typology model would better predict a panel’s decision to reverse a district court judgment than ideological mismatches based upon party of the appointing President between the panel and district court judge. Hypothesis: Stalwart and Regulating judges would exhibit a higher reversal rate when reviewing district court judges appointed by the President’s party opposite the majority of the panel when applying a deferential standard of review.

The typology model performed as predicted as to reversal rates using

deferential standards of review. As indicated in Figure 13, Stalwart, Steadfast and Error Correcting Judges (“Activist Types”) reversed district court judges using deferential standards of review at a rate 15% higher than other types of judges while they reversed using non-deferential standards at a nearly identical rate.

Although the 1.3% difference in reversal rates between activist and non-activist types might not seem that large, it accounts for thousands of cases every year. Certainly, other explanations such as the existence of “easy

tested cases in this Section were drawn from the circuits which regularly reported the identity of the district court judge.

10.0%8.7%

12.3% 12.0%

4.0%

6.0%

8.0%

10.0%

12.0%

14.0%

Activist Types Other Types

Pane

l Rev

ersa

l Rat

e

Judge Type

Figure 13 - Reversal Rates for Panels with Activist Types using Different Standards of Reviews

Deferential Non-Deferential


cases” and the constraints of substantive and procedural law ensured that judging styles still operated within certain limits. The increase in reversal rates due to judicial style was notable precisely because it has not been part of the traditional explanations for decisionmaking

Applying the ideology model, it was expected that higher reversal rates using deferential standards would occur when more panelists were appointed by the opposing President’s party than the district judge. There were four possible ideological configurations that a review of a district judge might entail: 1) The panel was ideologically united against the district judge; 2) two of the panelists were ideologically opposite of the district judge; 3) only one panelist was from the party opposing the district judge; and 4) the district judge was of the same appointing party as all of the panel judges. The ideological model predicted that the reversal rate using deferential standards of review would decline in the order of the scenarios listed above. However, as illustrated in Figure 14, the results did not support such a hypothesis.

It is unclear what should be made of the ideology model results. Strangely, the highest reversal rates using deferential and non-deferential standards occurred when the panelists were appointed by the same Presidential party. The lowest reversal rates occurred when the district judge was of the opposite ideology of two panelists. And the only one of the four scenarios where judges exhibited a high activist differential was when the district judge was aligned with two panelists. It is possible that some of these results could be due to the frequency of certain ideology combinations in different circuits (particularly for the first and fourth scenarios) where those circuits were more or less activist as a whole. Prior studies have also shown

11.8%

6.1%

9.0%

12.5%12.5%

7.8% 8.0%

13.6%

4.0%

6.0%

8.0%

10.0%

12.0%

14.0%

One on Three One on Two One on One Zero on Zero

Pane

l Rev

ersa

l Rat

e

Ideological Relationship of District Judge to Appellate Panelists

Figure 14 - Reversal Rates for Ideological Configurations with Different Standards of Reviews



how little ideology has mattered in district court decisionmaking which might further explain the seeming unrelated nature of district court ideology with panel ideology.95

The judge-type model predicted panels with partisan types would have reversed district court judges of the opposite party at a higher rate than those of the same party. Further, it was expected that the differential between same and opposite party district judges would be higher for partisan type judges than other judging types. Figure 15 indicates that there was evidence to support that hypothesis.

Regardless of the explanation, it appears that the judge type model was far more effective in predicting reversals using deferential standards than the ideology model.

Again, the results are not as dramatic as with the dissent predictions, but the data does support the role of judge type in forecasting partisan review of district judges. The presence of a single partisan judge on a panel doubled the difference between reversal rates using a deferential standard of review. In contrast, the ideological model still failed for the reasons described earlier in this Section.

B. Demographic and Biographical Characteristics In addition to considering the predictive power of judicial types, it was

also important to consider whether a judge’s style was related to demographic or biographical traits. For example, a judge who attended a

95 See, e.g., Gregory C. Sisk, Michael Heise & Andrew P. Morriss, Charting the

Influences on the Judicial Mind: An Empirical Study of Judicial Reasoning, 73 N.Y.U. L. REV. 1377, 1465 (1998)

2.0%

1.0%

0.2% 0.2%

0.0%

0.5%

1.0%

1.5%

2.0%

2.5%

Party Conflict Party Unity

Incr

ease

d Ra

te o

f Rev

ersa

ls fo

r Pa

nels

w

ith

Part

isan

Typ

es

Ideological Relationship between Panel Majority and District Court

Figure 15 - Reversal Rate Differential for Panels with Partisan Types featuring Different Standards of Review

and Ideological Relationships



particular law school, served first as a district judge, or was from a certain circuit might exhibit different decisionmaking behavior in statistically significant manner. Because there were 178 judges spread across nine judicial styles (an average of approximately 19.9 judges per style), the population sizes involved in various regression analyses were expected to find little or no evidence to reject the null hypothesis in most instances. And because the distribution among the categories was highly uneven, it is worth noting that future research based upon larger populations of judges might identify more correlations. Nonetheless, using multinomial logistic regression, some background traits were found to be correlated with certain judicial styles. Although each of the tested hypotheses is not stated below, they generally were structured as follows:

Hypothesis: [Background Characteristic] is correlated with the varied judicial categories.

1. Law School Attended

There has been little or no empirical study of whether a judge or

lawyer’s professional decisions were correlated with the law school that he or she attended. However, the data in this study showed that two significant correlations between a judge’s law school and his or her judicial type. First, the ranking of the law school attended (based upon the 2008 U.S. News and World Report law school rankings) was correlated with which judicial type a judge would belong. Figure 16 below indicates the average law school rank for judges in each category.

47

1820

2929

3538

42

0 10 20 30 40 50

SteadfastTrailblazing

StalwartMinimalistic

RegulatingPragmaticColleagial

Error CorrectingConsensus Building

US News & World Report Law School Ranking (2008)

Figure 16 - Average Law School Ranking by Judge Type (p=0.0369)


Perhaps most interesting is the judicial styles most strongly associated with ideology and partisanship (Steadfast and Trailblazing), on average, attended law schools that were much higher ranked (meaning they had a lower numerical ranking). In contrast, the types that epitomized collegiality (Consensus Building and Colleagues) had among the lowest average law school ranking. Although these results do not mean, for example, that a law graduate from Yale or Harvard would be destined to be a Trailblazer or Steadfast judge, some connection to law school education is clearly supported by the data.96

A second correlation was found based upon whether the law school attended was public or private.

97

96 p = 0.0369; pseudo R2 = 0.0243. Iteration 0: log likelihood = -337.30628; Iteration

1: log likelihood = -331.54943; Iteration 2: log likelihood = -330.15276; Iteration 3: log likelihood = -329.42114; Iteration 4: log likelihood = -329.13961; Iteration 5: log likelihood = -329.10242; Iteration 6: log likelihood = -329.10162; Iteration 7: log likelihood = -329.10162.

Figure 17 shows the breakdown of public and private law school graduates among the different judicial types.

97 p = 0.0319; pseudo R2 = 0.0249; Iteration 0: log likelihood = -337.30628; Iteration 1: log likelihood = -329.77443; Iteration 2: log likelihood = -329.15856; Iteration 3: log likelihood = -328.98971; Iteration 4: log likelihood = -328.92793; Iteration 5: log likelihood = -328.90523; Iteration 6: log likelihood = -328.89688; Iteration 7: log likelihood = -328.89381; Iteration 8: log likelihood = -328.89268; Iteration 9: log likelihood = -328.89226; Iteration 10: log likelihood = -328.89211; Iteration 11: log likelihood = -328.89206; Iteration 12: log likelihood = -328.89203; Iteration 13: log likelihood = -328.89203; Iteration 14: log likelihood = -328.89202; Iteration 15: log likelihood = -328.89202; Iteration 16: log likelihood = -328.89202; Iteration 17: log likelihood = -328.89202; Iteration 18: log likelihood = -328.89202; Iteration 19: log likelihood = -328.89202; Iteration 20: log likelihood = -328.89202; Iteration 21: log likelihood = -328.89202; Iteration 22: log likelihood = -328.89202; Iteration 23: log likelihood = -328.89202; Iteration 24: log likelihood = -328.89202; Iteration 25: log likelihood = -328.89202; Iteration 26: log likelihood = -328.89202; Iteration 27: log likelihood = -328.89202; Iteration 28: log likelihood = -328.89202; Iteration 29: log likelihood = -328.89202; Iteration 30: log likelihood = -328.89202; Iteration 31: log likelihood = -328.89202; Iteration 32: log likelihood = -328.89202; Iteration 33: log likelihood = -328.89202; Iteration 34: log likelihood = -328.89202.


A few categories stood out as having unusual distributions of private and public law school graduates. All five Steadfast judges attended private school. Over 78% of Pragmatic judges also graduated from private law schools. However, Trailblazing, Regulating, and Consensus Building judges more often matriculated from public law schools. Certainly, there was overlap between the law school ranking and the type of law school correlations. For example, the average ranking of a law school for Steadfast judges was approximately fourth overall. Because private law schools dominate the top of the rankings, it was, thus, unsurprisingly that Steadfast judges all attended private school. It was difficult to know the basis for this correlation without further exploration, but it further supported the proposition that where a judge attended law school is not just an insignificant biographical note in understanding their decisionmaking. 2. Appointing President’s Political Party

Although ideology was an important component of the judicial types,

only the absolute value of Ideology Score was used (yielding the Ideological Score). This meant that a strong conservative and liberal often were sorted into the same category. As a result, it did not necessarily follow that the party of the appointing President would correlate with the judicial type of the judge. Nonetheless, such a correlation was found based upon multinomial regression analysis of the data.98

98 p = 0.0075; pseudo R2 = 0.0309; Iteration 0: log likelihood = -337.30628; Iteration

1: log likelihood = -329.19248; Iteration 2: log likelihood = -327.32499; Iteration 3: log likelihood = -327.02647; Iteration 4: log likelihood = -326.93136; Iteration 5: log

Figure 18 below shows the

2 32

14

522

258 22

3 41

19

020

78 12

0%20%40%60%80%

100%Pe

rcen

tage

of J

udge

s

Figure 17 - Private versus Public Law School Attendance and Judicial Style (p=0.0319)

Private Public


breakdown among each judging style of which appointing President’s party nominated the judge.

Certain groups, such as Steadfast judges, demonstrated an ideological makeup outside of the normal distribution. In the case of Steadfast judges, their ranks were exclusively from those appointed by Democratic Presidents. Perhaps, given the relative unwavering nature of Steadfast judges, it was logical that they would be from the minority party in the federal judiciary. In contrast, over 78% of Pragmatic and 80% of Trailblazing judges were appointed by Republican Presidents. It was difficult to know if some of these trends would hold with a larger population of judges in certain groups, but regression did show statistical significance of the types collectively based upon appointing party. Notably, perhaps due to population sizes in some sub-groups, the specific appointing President was not correlated with the type of judge.99

likelihood = -326.89646; Iteration 6: log likelihood = -326.88362; Iteration 7: log likelihood = -326.8789; Iteration 8: log likelihood = -326.87716; Iteration 9: log likelihood = -326.87652; Iteration 10: log likelihood = -326.87629; Iteration 11: log likelihood = -326.8762; Iteration 12: log likelihood = -326.87617; Iteration 13: log likelihood = -326.87616; Iteration 14: log likelihood = -326.87615; Iteration 15: log likelihood = -326.87615; Iteration 16: log likelihood = -326.87615; Iteration 17: log likelihood = -326.87615; Iteration 18: log likelihood = -326.87615; Iteration 19: log likelihood = -326.87615; Iteration 20: log likelihood = -326.87615; =Iteration 21: log likelihood = -326.87615; Iteration 22: log likelihood = -326.87615.

99 p = 0.6084.

1 22

11

5

15 7 519

4 51

23

0

27 25 1115

0%20%40%60%80%

100%

Perc

enta

ge o

f Jud

ges

Figure 18 - Judge Type by Appointing President's Party (p=0.0075)

Democrat Republican


3. Senior Status A third background trait that was correlated with the judicial

categorization system was whether the judge had taken senior status at any time in or prior to 2008.100

Figure 19 below shows the distribution of judges who had taken senior status among the nine judicial styles.

100 p = 0.0076; pseudo R2 = 0.0309; Iteration 0: log likelihood = -337.30628; Iteration

1: log likelihood = -330.6079; Iteration 2: log likelihood = -327.82865; Iteration 3: log likelihood = -327.23075; Iteration 4: log likelihood = -327.01536; Iteration 5: log likelihood = -326.9363; Iteration 6: log likelihood = -326.90724; Iteration 7: log likelihood = -326.89656; Iteration 8: log likelihood = -326.89263; Iteration 9: log likelihood = -326.89118; Iteration 10: log likelihood = -326.89065; Iteration 11: log likelihood = -326.89045; Iteration 12: log likelihood = -326.89038; Iteration 13: log likelihood = -326.89036; Iteration 14: log likelihood = -326.89035; Iteration 15: log likelihood = -326.89034; Iteration 16: log likelihood = -326.89034; Iteration 17: log likelihood = -326.89034; Iteration 18: log likelihood = -326.89034; Iteration 19: log likelihood = -326.89034; Iteration 20: log likelihood = -326.89034; Iteration 21: log likelihood = -326.89034; Iteration 22: log likelihood = -326.89034; Iteration 23: log likelihood = -326.89034; Iteration 24: log likelihood = -326.89034; Iteration 25: log likelihood = -326.89034; Iteration 26: log likelihood = -326.89034; Iteration 27: log likelihood = -326.89034; Iteration 28: log likelihood = -326.89034; Iteration 29: log likelihood = -326.89034; Iteration 30: log likelihood = -326.89034; Iteration 31: log likelihood = -326.89034; Iteration 32: log likelihood = -326.89034; Iteration 33: log likelihood = -326.89034; Iteration 34: log likelihood = -326.89034; Iteration 35: log likelihood = -326.89034; Iteration 36: log likelihood = -326.89034; Iteration 37: log likelihood = -326.89034; Iteration 38: log likelihood = -326.89034; Iteration 39: log likelihood = -326.89034; Iteration 40: log likelihood = -326.89034; Iteration 41: log likelihood = -326.89034; Iteration 42: log likelihood = -326.89034; Iteration 43: log likelihood = -326.89034; Iteration 44: log likelihood = -326.89034; Iteration 45: log likelihood = -326.89034.

0 1

3

80 4 6

6 8

5 6

0

255 38 26

10 26

0%

20%

40%

60%

80%

100%

Perc

enta

ge o

f Jud

ges

Figure 19 - Senior Status and Judging Style (p=0.0076)

Senior Not Senior


A theoretical explanation for why certain judging types have higher representation of senior status judges than would be expected was not initially obvious. However, a judge’s self-selection offered one possible explanation. In order for a senior judge to appear in this study, he or she must have decided not to retire and still hear enough cases to qualify. Because senior judges can often retire and receive the same pay as they would by continuing to work, many of those that stayed on the bench were clearly not motivated by financial gain.101 Further, to hear a docket equivalent in size to active status judges was a decision made by the individual judge. It might follow, then, that the three Stalwart judges were understandably all senior status judges who were firmly committed to strong ideological and partisan positions. Of possible note, among senior status judges, the date at which they took senior status was uncorrelated with judge type.102

Any explanation for the correlation in this area needs further investigation.

4. Circuit The last background trait associated with judge type was the circuit on

which the judge was appointed.103

101 Stephen J. Choi, G. Mitu Gulati & Eric A. Posner, Are Judges Overpaid? A

Skeptical Response to the Judicial Salary Debate, 1 J. OF LEGAL ANALYSIS 47, 57-58 (2009).

Appendix D includes charts illustrating the distribution of judge types among each studied circuit. However, it is worth noting one prime example here. The Sixth Circuit, as illustrated in

102 p = 0.0722. 103 p < 0.0001; pseudo R2 = 0.2338. Iteration 0: log likelihood = -337.30628; Iteration

1: log likelihood = -323.05704; Iteration 2: log likelihood = -315.47043; Iteration 3: log likelihood = -301.68506; Iteration 4: log likelihood = -274.16192; Iteration 5: log likelihood = -261.57938; Iteration 6: log likelihood = -259.19609; Iteration 7: log likelihood = -258.68163; Iteration 8: log likelihood = -258.53413; Iteration 9: log likelihood = -258.48119; Iteration 10: log likelihood = -258.46172; Iteration 11: log likelihood = -258.45457; Iteration 12: log likelihood = -258.45193; Iteration 13: log likelihood = -258.45097; Iteration 14: log likelihood = -258.45061; Iteration 15: log likelihood = -258.45048; Iteration 16: log likelihood = -258.45043; Iteration 17: log likelihood = -258.45041; Iteration 18: log likelihood = -258.45041; Iteration 19: log likelihood = -258.4504; Iteration 20: log likelihood = -258.4504; Iteration 21: log likelihood = -258.4504; Iteration 22: log likelihood = -258.4504; Iteration 23: log likelihood = -258.4504; Iteration 24: log likelihood = -258.4504; Iteration 25: log likelihood = -258.4504; Iteration 26: log likelihood = -258.4504; Iteration 27: log likelihood = -258.4504; Iteration 28: log likelihood = -258.4504; Iteration 29: log likelihood = -258.4504; Iteration 30: log likelihood = -258.4504; Iteration 31: log likelihood = -258.4504; Iteration 32: log likelihood = -258.4504; Iteration 33: log likelihood = -258.4504; Iteration 34: log likelihood = -258.4504; Iteration 35: log likelihood = -258.4504; Iteration 36: log likelihood = -258.4504.


Figure 20 below has been marked by a much higher level of ideological division than the other Courts of Appeals.104

As measured by the Ideology Scores, of the 178 judges in the sample, an average Sixth Circuit judge had an Ideology Score that was 36.1 points from the mean for that circuit. The next most divided circuit was the Eighth, with an average variation of 19.3 points. Eight of the eleven circuits studied had average variations less than 11 points. Needless to say, the Sixth Circuit was an outlier among the circuits in terms of ideological divide. Such statistical findings were further supported by media and scholarly coverage of the opinions issued from that court.105

104 The chart was created using the 2008 Case Database and not the 2009 Testing

Database.

105 Dan Horn, 6th Circuit's Infighting Gets Personal, CIN. ENQ., Oct. 16, 2008, at 1A (“Political warfare over voter registration in Ohio spilled into federal court this week, reigniting a partisan fight on one of the nation's most powerful and divided courts…. The tone of the judges' written opinions was as notable as their stand on the issue: Some of them used politically-charged language, accused one another of ignoring the law and, in one case, questioned the integrity and motivations of fellow judges…. That kind of tough talk is unusual in most federal courts. But it's just the latest example of the internal strife that has afflicted the 6th Circuit for years as it wrestled with such issues as the death penalty, affirmative action and voter rights. ‘It's not a pretty sight,’ said Arthur Hellman, a University of Pittsburgh law professor who closely follows federal courts. ‘It casts doubt on the role of the courts as an institution apart from politics.’ All circuit courts, which are second only to the U.S. Supreme Court in their authority, are home to judges from different political and ideological backgrounds. At the 6th Circuit, however, those differences have led to several public disagreements, personal feuds and bitterly contested votes.”).

7.38.4

8.010.5

5.336.1

10.319.3

17.29.2

5.6

0 10 20 30 40

First (6)Second (15)

Third (13)Fourth (12)

Fifth (20)Sixth (16)

Seventh (14)Eighth (16)Ninth (36)Tenth (16)

Eleventh (14)

Average Ideology Score Variance

Circ

uit (

Qua

lifie

d Ju

dges

)Figure 20 - Average Ideological Variance by Circuit (p <

0.0001)


Based upon such sharp ideological differences, it was expected that the Sixth Circuit would have a disproportionate amount of the judging types associated with high ideology and independence. Figure 21 below includes a list of all of the Sixth Circuit judges in the 178 judge sample and their identified type (in order of judge type).

Figure 21 - Sixth Circuit Judging Types Judge Overall Type Eric L. Clay Trailblazing Liberal Democrat Danny J. Boggs Trailblazing Conservative Republican Richard A. Griffin Trailblazing Conservative Republican John M. Rogers Trailblazing Conservative Republican David W. McKeague Trailblazing Conservative Republican Gilber S. Merritt, Jr. Stalwart Liberal Democrat Jeffrey S. Sutton Regulating Moderate Republican Deborah L. Cook Regulating Conservative Republican Eugene E. Siler, Jr. Regulating Conservative Republican Alice M. Batchelder Regulating Conservative Republican Karen N. Moore Steadfast Liberal Democrat Ransey G. Cole, Jr. Steadfast Moderate Democrat Martha C. Daughtrey Steadfast Moderate Democrat Ronald L. Gilman Steadfast Conservative Democrat Julia S. Gibbons Minimalistic Moderate Republican Boyce F. Martin, Jr. Error Correcting Liberal Democrat

The Sixth Circuit had sixteen total judges with sufficient population sizes. Among those, there were five Trailblazing, one Stalwart, and four Steadfast judges. Thus, ten of the judges were of the most types prone to conflict. In the other ten circuits, there were only nineteen more judges of those three types. It is unclear if the judicial types in the Sixth Circuit were a partial product or cause of the ongoing situation there, but there is evidently a strong correlation. 5. Combined Predictive Factors

Among the four identified correlations above, there was most likely

substantial overlap. For example, graduates of specific law schools were not evenly distributed among the circuits. Similarly, senior judges had a different distribution of appointing President’s party than more recently appointed judges. As a result, it was helpful to reconsider the correlations when they were studied collectively. Combining law school ranking,


appointing President’s ideology, and a judge’s senior status106 yielded a very effective model for predicting judicial type out of the population of studied confirmed judges.107 Even without judicial circuit (a variable that was probably most removed from a judge’s chosen background factors), the model yielded very good results in predicting judicial behavior among the studied population.108 For example, using maximum likelihood analysis,109

106 Whether a law school was public or private was omitted from this analyses because

of its strong overlap with law school ranking.

107 p < 0.0001; pseudo R2 = .3253; Iteration 0: log likelihood = -337.30628; Iteration 1: log likelihood = -310.78885; Iteration 2: log likelihood = -305.36572; Iteration 3: log likelihood = -294.33663; Iteration 4: log likelihood = -276.64879; Iteration 5: log likelihood = -260.46749; Iteration 6: log likelihood = -244.48212; Iteration 7: log likelihood = -236.90805; Iteration 8: log likelihood = -232.27889; Iteration 9: log likelihood = -230.36757; Iteration 10: log likelihood = -229.78201; Iteration 11: log likelihood = -229.09448; Iteration 12: log likelihood = -228.15247; Iteration 13: log likelihood = -227.78679; Iteration 14: log likelihood = -227.66124; Iteration 15: log likelihood = -227.61616; Iteration 16: log likelihood = -227.59967; Iteration 17: log likelihood = -227.5936; Iteration 18: log likelihood = -227.59137; Iteration 19: log likelihood = -227.59055; Iteration 20: log likelihood = -227.59024; Iteration 21: log likelihood = -227.59013; Iteration 22: log likelihood = -227.59009; Iteration 23: log likelihood = -227.59008; Iteration 24: log likelihood = -227.59007; Iteration 25: log likelihood = -227.59007; Iteration 26: log likelihood = -227.59007; Iteration 27: log likelihood = -227.59007; Iteration 28: log likelihood = -227.59007; Iteration 29: log likelihood = -227.59007; Iteration 30: log likelihood = -227.59007; Iteration 31: log likelihood = -227.59007; Iteration 32: log likelihood = -227.59007; Iteration 33: log likelihood = -227.59007; Iteration 34: log likelihood = -227.59007; Iteration 35: log likelihood = -227.59007; Iteration 36: log likelihood = -227.59007; Iteration 37: log likelihood = -227.59007; Iteration 38: log likelihood = -227.59007.

108 p = 0.0001; pseudo R2 = 0.0864; Iteration 0: log likelihood = -337.30628; Iteration 1: log likelihood = -329.35013; Iteration 2: log likelihood = -311.6319; Iteration 3: log likelihood = -309.14297; Iteration 4: log likelihood = -308.43105; Iteration 5: log likelihood = -308.22145; Iteration 6: log likelihood = -308.17185; Iteration 7: log likelihood = -308.15653; Iteration 8: log likelihood = -308.15092; Iteration 9: log likelihood = -308.14885; Iteration 10: log likelihood = -308.14809; Iteration 11: log likelihood = -308.14781; Iteration 12: log likelihood = -308.14771; Iteration 13: log likelihood = -308.14767; Iteration 14: log likelihood = -308.14766; Iteration 15: log likelihood = -308.14765; Iteration 16: log likelihood = -308.14765; Iteration 17: log likelihood = -308.14765; Iteration 18: log likelihood = -308.14765; Iteration 19: log likelihood = -308.14765; Iteration 20: log likelihood = -308.14765; Iteration 21: log likelihood = -308.14765; Iteration 22: log likelihood = -308.14765; Iteration 23: log likelihood = -308.14765; Iteration 24: log likelihood = -308.14765; Iteration 25: log likelihood = -308.14765; Iteration 26: log likelihood = -308.14765; Iteration 27: log likelihood = -308.14765; Iteration 28: log likelihood = -308.14765; Iteration 29: log likelihood = -308.14765; Iteration 30: log likelihood = -308.14765; Iteration 31: log likelihood = -308.14765; Iteration 32: log likelihood = -308.14765; Iteration 33: log likelihood = -308.14765; Iteration 34: log likelihood = -308.14765; Iteration 35: log likelihood = -308.14765; Iteration 36: log likelihood = -308.14765.


it was determined that a graduate of the University of Chicago Law School who had not taken senior status appointed by a Democratic President would be in each of the nine judge types according to the probabilities in Figure 22 below.

However, if the judge were a Northwestern University School of Law

graduate who had not taken senior status and was appointed by a Republican President, Figure 23 includes the probabilities that he or she would exhibit a specific judging style:

109 Each of the results described in this section was completed using the spost add on to

Stata available here: http://www.indiana.edu/~jslsoc/spost.htm.

Trailblazing3% Consensus

Building3%

Regulating17%

Steadfast13%

Collegial22%Pragmatic

11%

Minimalistic8%

Error Correcting

23%

Figure 22 - Probability of Non-Senior University of Chicago Law School Graduate Appointed by Democratic

President of Exhibiting Judicial Styles


Although many of the probabilities between the two graphs are similar, one very significant difference concerns whether the judge would be Steadfast. The first judge has approximately 13% chance of being Steadfast, but the second judge has an essentially 0% probability. The second judge is also over twice as likely to be a Pragmatic judge.

Moving to another law school in Chicago that is not as highly ranked, a graduate of the Chicago-Kent College of Law who was appointed by a Democratic President and had not taken senior status would have the judge type probabilities in Figure 24.

Whereas the profiles of the Northwestern and Chicago law school

graduates had distributions across a wide range of judge types, the Chicago-

Trailblazing6%

Consensus Building

4%

Regulating21%Collegial

26%

Pragmatic24%

Minimalistic9%

Error Correcting10%

Figure 23 - Probability of Non-Senior Northwestern University School of Law Graudate Appointed by Republican President of Exhibiting Judicial Styles

Consensus Building

5%

Regulating16%

Collegial29%

Pragmatic10%

Minimalistic3%

Error Correcting

37%

Figure 24 - Probability of Non-Senior Chicago-Kent College of Law Graudate Appointed by Democratic

President of Exhibiting Judicial Styles


Kent alum has a very different set of likely outcomes. He or she is, among the various options, most likely to be either Error Correcting or Collegial and there is almost no probability that the judge will be Trailblazing, Stalwart, or Steadfast.

Of course, none of these maximum likelihood based probabilities indicate that a President must follow a simple biographical and demographic formula to pick the type of judge he or she is seeking. The studied population includes only those who were nominated, confirmed, and currently on the bench. The population of all potential nominees is much broader and surely not represented by the group studied herein. Further, although the background factor model gives insight into relevant factors that are related to judicial styles, there is nothing to support a causative connection in this study.

V. RELIABILITY, VALIDITY, AND LIMITATIONS Because the scope of this project was quite large, it is even more

essential than normal to consider the reliability and validity of the study. Making sure that this study is both repeatable and actually measured what it intended to are significant considerations in gauging the scholarly weight that it should be given. Further, the study needs to be understood within the boundaries of its methods, data, and analysis. Each of these concerns is discussed below.

A. Reliability

Reliability is the degree to which the measurement would yield the same

results when applied by others.110

Case data were downloaded from LexisNexis for each of the Circuits studied which provided a stable platform for replication by other researchers. Many of the variables were coded by computer software which should provide a high level of consistency among studies. The remaining data were coded by law students and recent law school graduates. On both

Because this study is the first to systematically identify judicial styles for judges on the Courts of Appeals, it is impossible to compare the results directly with other research. Instead, reliability can only be evaluated by the quality of the coding and analysis. Since this is not the first study using some version of this dataset, some of the discussion here is abbreviated.

110 Lee Epstein & Gary King, The Rules of Inference, 69 U. CHI. L. REV. 1, 83 (2002)

(“Reliability is the extent to which it is possible to replicate a measurement, reproducing the same value (regardless of whether it is the right one) on the same standard for the same subject at the same time.”).


the computer and human coded variables, quality checking was performed by other student and graduate coders as well as the Author. Accepted inter-coder reliability was achieved for every variable used in this study. In addition, a variety of checks were performed to ensure internal consistency of variables that were necessarily interconnected.111

B. Validity Validity is determined by whether the measurements used in an

empirical study reflects the concept(s) measured.112 Generally, validity can be understood along several dimensions. Professors Lee Epstein and Gary King previously identified three possible ways to view validity: “facial validity, unbiasedness, and efficiency.”113 Although establishing validity in each of those categories is not strictly required, 114

As Epstein and King stated, “[a] measure is facially valid if it is consistent with prior evidence, including all quantitative, qualitative, and even informal impressionistic evidence.”

this article considers each type in turn. Because validity of the four underlying judicial traits has been considered in prior scholarship, the validity discussion herein focuses solely on the nine judicial types. If there were validity problems with the underlying measures, however, those would certainly affect the validity of the results of this further study.

115

In the dataset, there are several judges that might be deemed to have sufficient notoriety such that qualitative assessments of their judicial types might be made. Figure 25 below includes a list of such judges, their component scores, and resultant type (including appointing President and ideological status based upon the Ideology Score):

Because this is the only study of its type using the described new data, it was necessary to focus exclusively on qualitative and impressionistic evidence of validity.

Figure 25 – Notable Judges

Judge Ideo. Act. Part. Ind. Overall Type Danny J. Boggs 40.1 20.7 34.7 64.0 Trailblazing

111 For example, the party labels in the coding include “criminal defendant.” Such a

party label precluded “civil plaintiff” or “civil defendant” from appearing in the outcome variable. Many cross-checks were employed to ensure quality and correct errors within the dataset.

112 Epstein & King, supra note 110, at 87. 113 Id. at 89. 114 Id. (“[N]o one of these is always necessary, and together they are not always

sufficient, even though together they are often helpful in understanding when a measure is more or less valid.”).

115 Epstein & King, supra note 110, at 89.


Conservative Republican

Deborah L. Cook 34.5 60.2 65.2 60.7 Regulating Conservative Republican

Frank H. Easterbrook 9.8 50.0 57.8 43.6 Pragmatic Moderate Republican

Edith H. Jones 15.6 51.0 21.3 51.5 Error Correcting Moderate Republican

Michael W. McConnell -3.2 29.4 45.1 58.5 Minimalistic Moderate Republican

Diarmuid F. O'Scannlain 45.2 38.2 55.1 33.4

Consensus Building Conservative Republican

Richard A. Posner 18.7 49.0 44.9 36.4 Pragmatic Moderate Republican

Stephen R. Reinhardt -20.4 61.2 24.6 72.4 Steadfast Liberal Democrat

Sonia Sotomayor -17.3 39.3 57.2 48.1 Regulating Moderate Democrat

Sidney R. Thomas -2.9 61.7 32.4 57.0 Error Correcting Moderate Democrat

Kim M. Wardlaw -17.5 36.8 71.8 46.2 Regulating Moderate Democrat

J. Harvie Wilkinson III 7.0 49.8 37.5 39.0 Collegial Moderate Republican

Ann C. Williams -12.5 60.8 35.2 38.5 Error Correcting Moderate Democrat

Diane P. Wood -28.0 48.9 53.5 26.5 Pragmatic Liberal Democrat


The judges listed above either have “celebrity” status as public intellectuals or judicial lightning rods, have been mentioned as possible Supreme Court Justices, or, in one case, actually been appointed to the Supreme Court.

Among the notable judges, there are a lot of good fits in terms of similarities between impressionistic and quantitative categorization. Judge Wilkinson, a judge who has long championed collegiality among judges is designated as a Collegial Moderate Republican.116 While some might be surprised at the “Moderate” categorization, his rebuke of what he saw as conservative judicial activism in the Supreme Court’s Heller117 opinion gives credence to the designation.118

Many of the notable judges sit on the Seventh Circuit including Judges Easterbrook, Posner, Williams, and Wood. Despite their occasional publicized disagreements,

119 there is nothing obvious in either Judge Posner or Easterbrook’s history that would conflict with the Pragmatic Moderate Republican grouping where they are both found. Judge Posner, in particular, is a good fit in the Pragmatist group even if his conception of judicial pragmatism is not exactly the same as discussed herein.120 Similarly, Judge Wood is a Pragmatic Liberal Democrat which coincides well with the impressions of her given when President Obama considered her for the Supreme Court.121 There are not as many public assessments of Judge Williams, Error Correcting Moderate Democrat seems to be an acceptable label.122

116 J. Harvie Wilkinson III, The Drawbacks of Growth in the Federal Judiciary, 43

EMORY L.J. 1147, 1170 (1994).

117 District of Columbia v. Heller, 128 S. Ct. 2783 (2008). 118 J. Harvie Wilkinson III, Of Guns, Abortions, and the Unraveling Rule of Law, 95

VA. L. REV. 253, 274, 322 (2009). 119 See, e.g., Ameet Sachdev, Blagojevich Jury Names Create Rift among Judges, CHI.

TRIB., Jul. 16, 2010, at C26 (“In an 18-page opinion, Judge Richard Posner rebuked the July 2 decision written by Chief Judge Frank Easterbrook. The two jurists rarely clash over legal reasoning, so Posner's strongly worded critique exposed a level of dissension over this case that has lawyers buzzing.”).

120 See generally, Michael Sullivan & Daniel J. Solove, Can Pragmatism Be Radical? Richard Posner and Legal Pragmatism, 113 YALE L.J. 687 (2003).

121 See, e.g., Cross-Court Shot, INVESTOR'S BUSINESS DAILY, Apr. 29, 2010, at A10 (“Anthony Kennedy may be the swingiest of Supreme Court swing votes in U.S. history, and President Obama has made it clear that the fact is not lost on him. Reports say he is considering a liberal activist who could sway him, such as 7th Circuit Judge Diane Wood. Wood has had some influence on conservatives -- like her court's chief judge, Frank Easterbrook, and fellow judge Richard Posner.”).

122 Christi Parsons & Bob Secter, 2nd Chicago Judge in Mix for High Court; Ex-

Federal Prosecutor Was Named to Bench by Reagan, CHIC. TRIB., Apr. 22, 2010, at C7 (“A second federal appellate judge in Chicago, Ann Claire Williams, has been added to the


Justice, then Judge, Sotomayor was considered a Regulating Moderate Democrat. Because she was a Supreme Court nominee, the public record concerning her judging qualities is necessarily filled with hyperbolic vitriol123 and praise.124 Nonetheless, her background (having been originally been nominated to the district court by President H.W. Bush) and general record on the bench are indicators of moderation.125

Judge O’Scannlain is considered a Consensus Building Conservative Republican. Ninth Circuit court watchers would probably not quibble with the “Conservative” label, but might be surprised by the Consensus Building designation.

And her designation as a Regulating judge befits a judge who at least has some reasonable partisan applications of the law that might have provided some basis for objecting to her appointment to the Supreme Court.

126 However, as the results of this study have indicated, there is no strong connection between exhibiting higher levels of ideology and not drawing or engaging in dissents. There is nothing in the public record which would discount Judge O’Scannlain as a Consensus Builder. Two of Judge O’Scannlain’s foils on the Ninth Circuit, Judges Reinhardt and Wardlaw, were typed as a Steadfast Liberal Democrat and Regulating Moderate Democrat, respectively. Judge Reinhardt’s classification fits perfectly with his public reputation.127 Similarly, Judge Wardlaw is often considered quite liberal on some issues, but conservative on others.128

roster of potential nominees to fill the upcoming vacancy on the U.S. Supreme Court, the White House confirmed Wednesday.... Williams brings another potential advantage to the table in an era when any Obama nomination risks being ground up in the partisan divide of Washington: She was first named to the bench in 1985 by Republican President Ronald Reagan.... Williams has described herself as a political independent.”).

Because she is

123 See, e.g., Tom LoBianco, Lawyers Tag Sotomayor as “Terror on the Bench”, WASH. TIMES, May 29, 2009, at A01.

124 Robert Barnes, Battle Lines Are Drawn on Sotomayor Nomination, WASH. POST, May 28, 2009, at A1.

125 Charlie Savage, Uncertain Evidence for ‘Activist’ Label on Sotomayor, N.Y. TIMES, June 20, 2009, at A10.

126 Bob Egelko, Court to Review Recall Ruling; Makeup of 11-Judge Panel Suggests Oct. 7 Election Likely to be Approved, S.F. CHRON., Sep. 20, 2003, at A1.

127 David G. Savage, Did Victim's Photo Prejudice a Jury?; Another Ruling by the Liberal-Leaning 9th Circuit Comes under Supreme Court Review, L.A. TIMES, Sep. 18, 2006, at A1.

128 Howard Mintz, Two California Latino judges emerge as candidates for Supreme Court, SAN JOSE MERCURY NEWS, May 16, 2009, at A1 (“On the 9th Circuit, Wardlaw has been considered a moderate, more liberal on social issues such as civil rights but at the center of the court on criminal matters, including penning the decision rejecting the final appeals of California death-row inmate Clarence Ray Allen before he was executed. Even conservatives concede they'd have trouble contesting Wardlaw's qualifications. ‘She would be pretty bulletproof,’ said Curt Levey, head of the conservative Committee for Justice, which is vetting all of the potential nominees for a possible Republican political fight.


Regulating judge, she does exhibit a high level of partisanship in reviewing lower court judgments. So, it is possible that her liberal nature primarily surfaces in her interactions with the district courts, but her overall profile is moderate by virtue of her combined liberal and conservative views. Judge Thomas, another Ninth Circuit judge, was a relative unknown until some progressive groups championed him as a potential nominee to the Supreme Court.129 That seems to be an odd fit with his type, Error Correcting Moderate Democrat. However, many commentators noted that there was nothing particularly ideologically liberal in Judge Thomas’ record other than a couple timely high profile decisions which caught the attention of progressive advocacy groups.130

Judges Cook and Jones were both floated as potential nominees for the Supreme Court by President George W. Bush.

131 Judge Cook’s categorization as a Regulating Conservative Republican would explain why she would be a popular choice in some conservative circles.132 In contrast, Judge Jones is listed as an Error Correcting Moderate Republican, but her reputation is generally of a strong conservative judge.133

Judge Boggs, of the Sixth Circuit, has been at the forefront of many of the judicial battles on that Circuit.

However, the public record is limited on Judge Jones and it is unclear if the type given is far off the mark.

134 Viewing Judge Boggs as a Trailblazing Conservative Republican certainly fits with his role as a strong conservative voice in the federal judiciary. Judge McConnell, who was a law professor, became a federal judge, and is now a law school dean, has been a vocal supporter of a philosophy of judicial restraint.135

Overall, the categorizations represent fairly reasonable assessments of the notable judges based upon public impressions of the judges. There are, however, some designations with which some might disagree. Nonetheless,

Thus, his categorization as a Minimalistic Moderate Republican is an appropriate designation.

‘She’s not a fire-breathing liberal. But that's not to say we wouldn't try to use it as a 'teaching moment.’”).

129 John Schwartz, Long Shot for Court Has Reputation for Compassion and Persuasion, N.Y. TIMES, May 6, 2010, at A17.

130 Id. 131 Renee Montagne, Possible Supreme Court Nominees, NATIONAL PUBLIC RADIO,

Jul. 8, 2005 (noting that Judge Jones had made the shortlist, but Judge Cook had not). 132 Id. (noting Judge Cook’s “firmly conservative credentials”). 133 Peter Wallsten & Richard B. Schmitt, Vacancy on the Supreme Court, L.A. TIMES,

Jul. 2, 2005, at A03. 134 David Hawpe, Clearing the Air on Sixth Circuit Judges' Junkets and Jousts, THE

COURIER-JOURNAL, Oct. 1, 2006, at 2H. 135 Emily Bazelon & David Newman, The Supreme Court Shortlist, SLATE.COM, Jul. 1,

2005.


that is, in part, the value of the typology system. If the categories merely confirmed what was already known, it would not add as much to our collective understanding of decisionmaking. That the categorizations provide an acceptable level of agreement with qualitative evidence for the highest profile judges provides facial validity for the study.

Epstein and King wrote that, “[a] measurement procedure is unbiased if it produces measures that are right on average across repeated applications.”136

The last validity concern is “the basic idea being to choose the one with the minimum variance.”

There is an area where the methodology of this study fairs very well. Relying on computer coding, quantitative analysis, and excluding judges with insufficient population sizes provides a stable, unbiased measurement methodology. There is some potential for bias in the decision as to the appropriate number of judicial types, but that is inevitable in the use of cluster analysis.

137

Treating the judges studied in 2008 as the population instead of a sample of a larger body of cases diminishes the level of variance in the study. Further, the data was analyzed with frequentist statistics which have no variance outside of a sampling structure.

C. Limitations There are several limitations to the data used and results included in this

study. As a result, it is essential to be absolutely clear about the proper scope of inferences that should be drawn herein and in future research. Further, relying on a new dataset and data gathering techniques gives greater reasons to be cautious on the interpretation of this study.

1. Time Limitations

All the data studied in this Article were from 2008 opinions. The only

use of other data was in the testing phase (which used 2009 data) and not in the typology methodology. As a result, the labels given to judges in this article are only representative of that time period. Studies in other areas have shown the potential for drift over time which is not accounted for by the limited time frame used here.138

136 Epstein & King, supra note

Further, the population of 178 judges examined here is not representative of judges from other time periods. This

110, at 92 (emphasis removed). 137 Id., at 95. 138 See Lee Epstein, Andrew D. Martin, Kevin M. Quinn, & Jeffrey A. Segal,

Ideological Drift Among Supreme Court Justices: Who, When, and How Important?, 101 NW. U. L. REV. 1483 (2007) (finding in their study that an ideological drift among Supreme Court Justices occurred over time).


group was largely composed of nominees of Presidents Clinton and George W. Bush. And the nominees from older Presidents were not necessarily a representative sample of all nominees by those Presidents. Particularly in this era where there are enormous political fights about nominations to the Courts of Appeals, it is unclear if any results here can be generalized across time. Lastly, even assuming that the nine styles described herein account for the extent of modern judge types, there is no reason to believe that those styles encapsulate the possible approaches of judges used throughout time.

2. Data Gathering and Coding Limitations

The Case Database excluded opinions which did not use language

relevant to a standard of review when they were originally downloaded.139

A related limitation concerns the use of case issue codes. Although the automated coding and supplemental human coding process expanded the scope of case issues tested, there were often limited numbers of cases with in some areas. In other case areas, there were significant geographic skews in where the cases were heard. As a result, it was difficult to know which case issue mix variations would alter the underlying component scores and cluster analysis. Ultimately, the only case issue mix that showed repeated differences across circuits was the balance between civil and criminal cases. If further coding were done which provided more representative data for varied case issues which distorted the judge scores, the results here would have to be adjusted accordingly. So, any inferences here should be limited by the degree of knowledge about the effects of case issue mixes on judicial behavior.

The remaining 2008 opinions of the 178 judges examined in this study were not considered. If this study were generalizing about those judge’s entire 2008 performance, the sampling technique used would not have ensured randomness or representativeness. As a result, the study is limited to the population of cases outlined earlier and no inferences should be made as to cases outside of the dataset. Further, the data included neither the D.C. Circuit nor Federal Circuit. It is unknown whether judges from those courts would fit under the typology described in this Article.

139 Because the original dataset was created to implement the measure of judicial

activism described in this Article, there was an emphasis on standard of review language in accumulating the data. Further, because of limited research resources and uncertain model efficacy at the outset, coding was limited to certain fields. While more fields were coded at a later date, some of the initial constraints on the data remained. Hopefully, with the progression of automated case coding, the results here can be tested with larger populations.


3. Selection Effects Because this is a study of the federal appellate courts, there is an

inevitable concern about selection effects.140 A selection effect occurs when a party other than those being studied (e.g., a litigant) makes a decision (e.g., settlement) which is not accounted for by the study methods, but should be.141

However, the significance of selection effects in this study is more limited than other analyses of the Courts of Appeals because of the scope of the project. The underlying measures used to create the judicial typology are all relative in nature. They do not attempt to state that a judge is, for example, “liberal” by some external objective scale. Instead, the judicial type model merely states that relative to the other judges studied, one is, for example, more Steadfast in their approach. As a result of this relativistic model, the key concern with selection effects is whether they are relatively equal across the judge population. As long as judges have comparable selection effects (or the study accounts for uneven selection effects in some way), the selection effects should not limit the inferences drawn from the study. Nonetheless, it is helpful to appreciate the ways that selection effects might complicate the judicial typology.

There were various moments when a selection effect could have occurred in the data for this study including: pre-filing, pretrial, during trial, pre-verdict, post-verdict, pre-appeal, during appeal, and post-appeal.

The classic model of selection effects relies on the incentives of litigants. The Priest-Klein hypothesis predicted that, if parties had perfect information, case outcomes would split evenly (50%) between affirmances and reversals since the parties would settle as needed to avoid other outcomes.142

140 Keith N. Hylton, Asymmetric Information and the Selection of Disputes for

Litigation, 22 J. LEGAL STUD. 187 (1993); George L. Priest & Benjamin Klein, The Selection of Disputes for Litigation, 13 J. LEGAL STUD. 1 (1984); Cass R. Sunstein, Judging National Security Post-9/11: An Empirical Investigation, 2008 SUP. CT. REV. 269, 271.

This study, and consistent with prior examinations, provides ample reason to doubt the simple Priest-Klein hypothesis because the rate of affirmances is far higher than 50% (regardless of standard of review used) in civil and criminal cases. Notably, the Priest-Klein model was limited to

141 Kate Stith, The Risk of Legal Error in Criminal Cases: Some Consequences of the Asymmetry in the Right to Appeal, 57 U. CHI. L. REV. 1, 19 n.55 (1990).

142 Priest & Klein, supra note 140, at 4-5. Numerous studies have examined whether empirical evidence supports the Priest-Klein hypothesis. See, e.g., Theodore Eisenberg, Testing the Selection Effect: A New Theoretical Framework with Empirical Tests, 19 J. LEGAL STUD. 337 (1990); Randall S. Thomas & Kenneth J. Martin, Litigating Challenges to Executive Pay: An Exercise in Futility?, 79 WASH. U. L. Q. 569, 590-91 (2001); Robert E. Thomas, The Trial Selection Hypothesis Without the 50 Percent Rule: Some Experimental Evidence, 24 J. LEGAL STUD. 209, 222-26 (1995).


civil cases because the settlement structure, plea bargaining, in criminal cases creates radically different incentives such that nothing close to a 50% affirmance rate is exhibited (even with the bar of a certificate of appealability at the federal level).

Beyond the traditional Priest-Klein model, there is still a substantial area of selection effects. However, at the federal appellate level, the marginal cost of an appeal is very low compared to that of a trial.143 This means that the chance of settlement before the appeal is considered to be quite low. Uncertainty about appellate outcomes also creates an environment where parties cannot reach agreement about the potential risks and costs. The legal issues are usually close in civil cases144 and the parties do not know which judges will decide the appellate case until right before oral argument.145 By that time, the briefing has been completed,146

The typology ultimately accounted for the differing selection effects in criminal and civil cases by adjusting individual judge scores to assume each judge had the average proportion of criminal and civil cases. Beyond that, it is unclear if there are any uneven selection effects among the judge population. To the degree that the parties know the substantive and procedural law of a circuit and make the same basic risk assessments (which is what a selection effects model would generally predict), one might expect that regardless of differences among the circuits and judges, that the projected odds in individual cases would be similar. As a result, while selection effects inherently limit such studies, it is difficult to discern if such effects had a prominent role in shaping the data and results of this study.

settlement is unrealistic, and the marginal cost of oral argument is low.

143 See Meehan Rasch, Not Taking Frivolity Lightly: Circuit Variance in Determining

Frivolous Appeals Under Federal Rule of Appellate Procedure 38, 62 ARK. L. REV. 249, 264 (2009).

144 Brian Z. Tamanaha, The Distorting Slant in Quantitative Studies of Judging, 50 B.C. L. REV. 685, 748 (2009) (“As the authors acknowledge, the subset of cases that are actually appealed following trial are more likely to have ‘a degree of indeterminacy in the law.’” (quoting CASS R. SUNSTEIN ET AL., ARE JUDGES POLITICAL?: AN EMPIRICAL ANALYSIS OF THE FEDERAL JUDICIARY 16 n.20 (2006))).

145 Richard L. Revesz, Litigation and Settlement in the Federal Appellate Courts: Impact of Panel Selection Procedures on Ideologically Divided Courts, 29 J. LEGAL STUD. 685, 688 (2000) (“With one exception, the United States Courts of Appeals announce the composition of their panels only shortly before the oral argument, typically after all the briefs have been filed.”).

146 See id.


CONCLUSION Judges are like anyone else – they differ in their approach at work, can

be assessed in a multitude of ways, and do not perform in isolation. This study provides one way to understand how the judges on the Courts of Appeals. The advantages of the typology are not just in the predictive value of it compared to traditional methods. The nine judge categories illustrate that our vocabulary and understanding of judicial decisionmaking has grown stagnant. By limiting the scope of our inquiry and methods, we have missed tremendous opportunities to provide valuable quantitative information about some of the most important actors in our constitutional system. Hopefully, regardless of whether this particular typology of judging styles stands the test of time, it has opened up possibilities for new avenues of exploration.

* * *


APPENDIX A – COMMON ADJUSTMENT METHODOLOGY

As noted in the discussion of the four scores, there were several common adjustments made to the raw scores to incorporate specific factors that might explain portions of the underlying variance measured by the scores. The four possible common adjustments were for a judge’s circuit (applied to Ideology, Activism, and Partisanship Scores), “super-panel” effects (applied to Ideology, Activism, and Partisanship Scores, case issue mix (applied to all four measures), and common scaling (applied to all four measures).

The circuit adjustment relied on two basic premises: 1) a variety of differences between the circuits might explain variance in behavior; and 2) judges traveling between circuits would provide a means to calculate the amount of variance that was due to circuit differences. The first premise was likely to be non-controversial, but it should be noted that the adjustments made to the Activism and Partisanship Scores were very minor because of limited differences in behavior observed by traveling judges. The second contention concerning senior status judges sitting on panels in different circuits warrants further explanation.

There were 26 senior status judges who were on panels that issued opinions in 2008 in more than one circuit. Combined, they accounted for votes on 2,482 panels in the Case Database. Because each of those panels afforded up to three possible interactions with other judges (two for co-panelists and one for the district judge), there were potentially nearly 7,500 data points derived from these judges. The assumption was made that whatever trait was studied would remain constant in any circuit that the judge sat. So, when Judge Pasco Bowman heard cases in the Eighth and Eleventh Circuits, he would have the same activism, ideology, and partisanship traits. Thus, any change in his behavior between the two circuits would be due to unidentified differences in the circuits.

Because the distribution of the traveling votes was not evenly spread among the circuits, the 26 judges had their collective scores aggregated weighted according to how many votes each one issued in a particular circuit. Doing that allowed an expected traveling judge value to be determined for each circuit. That was computed using this formula (Ideology is used in the example, but the principal is the same for each score):

Expected Ideology Score = (Judge A votes in Circuit X * Judge A Ideology Score + Judge B votes in Circuit X * Judge B Ideology Score + Judge C votes in Circuit X * Judge C Ideology Score) / (Judge A votes in Circuit X + Judge B votes in Circuit X + Judge C votes in Circuit X)


The actual scores were then computed collectively (as one hypothetical traveling judge) for all of the traveling judges within a circuit.147 The difference between the expected and actual scores was the basis for the circuit adjustment. The raw adjustment number was then divided by the number of traveling votes in the circuit so that the circuit differential could be determined on a per-vote basis. Then each judge within the circuit would have the circuit adjustment applied based upon the number of votes he or she issued.148

“Super-panel” effects were the expected variance in voting caused by the political background of co-panelists and district judge being reviewed. Unlike the results in other panel effects studies, the integration of the district judge created new values which warranted different adjustments. It was determined that the party of the appointing President on the “super-panel” affected individual judge’s decisions to dissent and reverse. Figures 26 and 27 below indicate the magnitude of those differences based upon the results from the Case Database.

147 The collective treatment was traveling judges was necessary because individual

judges often had too few votes in particular circuits. As a result, combining the scores was the only means to ensure that the majority of the traveling judge information was not discounted entirely. If more data were available, a more complex adjustment system could be utilized.

148 Prior versions of the Ideology and Activism Scores were computed using circuit adjustments that were made on a per-judge basis. The switch to a per-vote basis explains a substantial portion of the changes in the Scores used in this study.

1.11%

1.38%

1.05%

0.0%0.2%0.4%0.6%0.8%1.0%1.2%1.4%

All Same Party Panel Same, District Different

Split Party Panel

Dis

agre

emen

t Ra

te w

ith

Co-P

anel

ists

Party of Appointing President of Co-Panelists and District Judge

Figure 26 - Panel Disagreement Rate by Appointing President's Party of Co-Panelists and District Judge


Consequently, totals were tallied for each judge based upon the six

possible political configurations for a judge’s two co-panelists and district court judge being reviewed. Those potential arrangements were: 1) two Republican co-panelists, Republican district judge (“RR-R”); 2) two Republican co-panelists, Democratic district judge (“RR-D”; 3) one Republican and one Democratic co-panelist, Republican district judge (“RD-R”; 4) one Republican and one Democratic co-panelist, Democrat district judge (“RD-D”); 5) two Democratic co-panelists, Republican district judge (“DD-R”); and 6) two Democratic co-panelists, Democratic district judge (“DD-D”). Based upon the data in Figures 26 and 27, it was determined how much variance in the judge scores could be explained by sitting on panels with the above six configurations. For example, if a judge sat on 30 RR-R panels and 10 DD-D panels, it was assumed that the 10 DD-D panels would cancel out the effects of 10 RR-R panels. However, the remaining 20 RR-R panels would pull the studied judge in a conservative direction and lower his or her reversal rate. As a result, the judge’s respective scores were adjusted based upon the expected magnitudes of those effects.

The Courts of Appeals studied here reviewed different sets of cases from different sets of district judges based upon geography.149

149 With the exception of the Federal Circuit, which was not included in this study, the

remaining circuits have geographic and not subject-matter based jurisdiction. Paul R. Michel, Foreword, Assuring Consistency and Uniformity of Precedent and Legal Doctrine in the Areas of Subject Matter Jurisdiction Entrusted Exclusively to the U.S. Court of Appeals for the Federal Circuit: A View from the Top, 58 AM. U. L. REV. 699, 702 (2009).

Within each

20.5%

22.9%

20.6%

19.0%19.5%20.0%20.5%21.0%21.5%22.0%22.5%23.0%23.5%

All Same Party Panel Same, District Different

Split Party Panel

Reve

rsal

Rat

e

Party of Appointing President of Co-Panelists and District Judge

Figure 27 - Reversal Rate by Appointing President's Party of Co-Panelists and District Judge


circuit, each judge was randomly assigned to hear cases usually on panels of three judges. As a result, any given judge studied could have had a vastly different mix of issues to rule upon than the average judge. Consequently, an adjustment was made based on differences in dockets of judges in order to decrease the degree to which any unobserved variables would affect the results.150

The last common adjustment was much simpler. Each score was adjusted to a similar scale. There were 178 studied judges who had at least 200 interactions with other judges. Scaling was done so that those 178 judges would define the extent of the scales. For the Activism, Partisanship, and Independence Scores were placed from 0 to 100 and the Ideology Score was scaled from -100 to 100. In the former case, this was done by moving the lowest observed raw score to the 0 and then multiplying all of the values by a constant needed for the highest score to reach 100. In the latter case (Ideology), the multiplier chosen was the value that would either move the highest score to 100 or lowest score to 100. It was determined that there was no reason to force symmetry by making one judge have a score of -100 and another 100.

In the Case Database, one particular distinction in case types proved significant in both dissent and reversal rates. In criminal cases, judges on panels agreed 99.1% of the time and affirmed the judgment of the lower court in 83.2% of cases. In contrast, the panel disagreement rate in civil cases was 98.5% and the reversal rate was 73.3%. There were also intersecting differences when deferential and non-deferential standards of review were applied in civil and criminal cases. As a result, each judge’s criminal and civil cases were grouped and had scores calculated separately. When sample sizes for individual subcategories were too low (i.e., deferential standard criminal cases), then the scores were interpolated based upon expected and actual scores with that judge’s case mix. The criminal and civil scores were then weighted according to the distribution of the average judge.

150 See David C. Vladeck, Keeping Score: The Utility of Empirical Measurements in

Judicial Selection, 32 FLA. ST. U. L. REV. 1415, 1433-34 (2005) (discussing, in the context of the Choi and Gulati study, the need to account for differences among circuit caseloads in creating empirical measures).


APPENDIX B – COMPONENT SCORES OF JUDGES The following table includes Ideology, Activism, Partisanship, and Independence Scores for judges with at least 200 interactions with other judges. Judge Cir. N151 Id. Act. Part. Ind. Ambro, Thomas L. 3 398 4.4 70.4 42.1 48.0 Anderson, Robert L., III 11 1007 0.4 47.0 43.1 35.4 Anderson, Stephen H. 10 332 7.0 39.8 67.5 19.7 Baldock, Bobby R. 10 352 5.1 44.8 45.0 53.9 Barkett, Rosemary 11 949 2.5 56.5 45.4 41.7 Barksdale, Rhesa H. 5 678 20.4 54.1 58.1 31.7 Barry, Maryanne T. 3 387 0.5 73.2 25.7 31.1 Batchelder, Alice M. 6 305 54.9 63.7 53.5 66.7 Bauer, William J. 7 389 8.9 67.2 47.5 47.0 Bea, Carlos T. 9 393 9.5 93.7 76.6 49.8 Beam, Clarence A. 8 276 36.3 47.4 50.6 56.2 Beezer, Robert R. 9 219 6.3 46.2 35.2 28.5 Benavides, Fortunato P. 5 717 1.6 65.9 41.2 26.9 Benton, William D. 8 659 5.5 33.7 48.5 46.8 Berzon, Marsha S. 9 370 47.7 33.2 24.1 21.2 Birch, Stanley F., Jr. 11 1005 9.6 36.9 45.0 36.0 Black, Susan H. 11 1044 0.4 53.5 37.4 33.6 Boggs, Danny J. 6 235 40.1 20.7 34.7 64.0 Boudin, Michael 1 233 1.7 53.0 61.7 53.8 Bowman, Pasco M., II 8 201 15.8 86.8 49.5 25.2 Bright, Myron H. 8 217 72.9 100.0 69.1 0.0 Briscoe, Mary B. 10 501 10.8 26.4 40.2 67.6 Brorby, Wade 10 230 2.2 42.1 59.0 23.2 Bybee, Jay S. 9 357 2.9 61.8 74.7 55.8 Bye, Kermit E. 8 559 3.8 56.2 43.2 46.2 Cabranes, Jose A. 2 474 6.6 62.2 29.6 24.4

151 N refers to the number of interactions the listed judge had with other judges (district

and appellate) in the dataset.


Calabresi, Guido 2 267 14.8 34.4 51.7 26.9 Callahan, Consuelo M. 9 372 4.3 65.2 36.7 41.9 Canby, William C., Jr. 9 324 13.3 67.6 0.0 50.5 Carnes, Edward E. 11 1028 2.5 55.6 48.1 34.4 Chagares, Michael A. 3 363 8.4 55.5 66.7 36.7 Clay, Eric L. 6 379 30.4 36.2 20.1 65.9 Clement, Edith B. 5 693 0.0 52.5 57.9 37.2 Clifton, Richard R. 9 306 10.3 66.9 65.3 47.1 Cole, Ransey G., Jr. 6 419 13.0 62.4 46.8 75.7 Colloton, Steven M. 8 591 2.6 63.3 42.6 54.9 Cook, Deborah L. 6 266 34.5 60.2 65.2 60.7 Cudahy, Richard D. 7 221 13.8 69.8 69.6 54.7 Daughtrey, Martha C. 6 313 7.0 53.4 39.1 80.0 Davis, W. Eugene 5 672 6.2 37.8 51.0 37.6 Demoss, Harold R., Jr. 5 267 5.1 44.5 33.6 65.3 Dennis, James L. 5 702 0.4 48.2 44.6 46.2 Dubina, Joel F. 11 1020 5.2 54.2 55.8 33.0 Duncan, Allyson K. 4 633 0.4 53.7 26.4 51.4 Easterbrook, Frank H. 7 369 9.8 50.0 57.8 43.6 Ebel, David M. 10 381 2.9 27.1 30.3 77.3 Edmondson, James L. 11 384 9.6 40.8 53.7 53.4 Elrod, Jennifer W. 5 534 1.1 68.3 43.9 33.0 Evans, Terence T. 7 483 3.3 72.6 43.8 53.1 Fay, Peter T. 11 408 11.0 38.7 49.1 38.0 Fernandez, Ferdinand F. 9 246 42.3 86.4 100.0 39.5 Fisher, D. Michael 3 448 9.0 61.0 81.1 42.5 Fisher, Raymond C. 9 426 12.5 31.1 50.6 55.9 Flaum, Joel M. 7 480 7.3 47.0 70.7 31.5 Fletcher, Betty B. 9 416 8.6 46.3 64.9 67.9 Fletcher, William A. 9 323 22.8 22.6 30.9 29.7 Fuentes, Julio M. 3 350 7.3 73.3 39.3 32.9 Garza, Emilio M. 5 672 16.9 43.1 34.2 39.1 Gibbons, Julia S. 6 363 3.2 35.5 33.9 62.3 Gibson, John R. 8 344 42.6 49.0 31.1 14.0 Gilman, Ronald L. 6 411 23.9 77.1 45.0 79.9


Gorsuch, Neil M. 10 430 3.9 45.6 56.2 72.3 Gould, Ronald M. 9 395 24.8 57.7 54.0 56.7 Graber, Susan 9 438 44.0 48.7 35.0 35.2 Gregory, Roger L. 4 663 10.7 51.9 49.8 65.6 Griffin, Richard A. 6 425 56.5 36.8 30.6 59.6 Gruender, Raymond W. 8 598 46.9 41.4 50.9 32.1 Hall, Cynthia H. 9 225 5.9 58.5 59.3 30.7 Hall, Peter W. 2 451 9.5 36.1 28.3 20.9 Hamilton, Clyde H. 4 495 0.3 70.5 46.2 46.7 Hansen, David R. 8 213 6.6 33.0 28.8 61.7 Hardiman, Thomas M. 3 359 22.6 61.6 65.2 33.2 Hartz, Harris L. 10 454 11.6 23.2 36.5 69.2 Hawkins, Michael D. 9 369 22.1 42.9 49.7 31.4 Haynes, Catharina 5 339 2.8 46.2 17.6 35.4 Higginbotham, Patrick E. 5 615 4.1 55.0 36.6 32.0 Holmes, Jerome A. 10 453 6.7 38.9 35.4 32.3 Howard, Jeffrey R. 1 340 1.2 64.6 76.5 46.2 Hull, Frank M. 11 1053 0.5 45.5 38.3 30.3 Ikuta, Sandra S. 9 350 28.9 53.7 60.7 44.0 Jacobs, Dennis G. 2 336 6.3 53.0 39.5 43.5 Jolly, E. Grady 5 689 10.4 47.6 35.1 43.5 Jones, Edith H. 5 437 15.6 51.0 21.3 51.5 Jordan, Kent A. 3 432 2.2 47.4 60.1 52.1 Kanne, Michael S. 7 420 4.0 48.3 51.4 41.5 Katzmann, Robert A. 2 354 4.7 52.9 28.0 36.9 Kelly, Paul J., Jr. 10 492 4.8 58.2 58.7 59.9 King, Carolyn D. 5 737 4.8 43.3 45.7 47.0 King, Robert B. 4 699 2.5 54.9 37.6 36.2 Kleinfeld, Andrew J. 9 285 25.9 65.2 69.6 15.5 Kravitch, Phyllis A. 11 384 6.3 36.5 47.9 37.5 Leavy, Edward 9 264 3.7 64.0 41.3 32.2 Lipez, Kermit V. 1 300 2.5 78.7 20.4 51.0 Livingston, Debra A. 2 419 6.7 52.3 46.2 34.0 Loken, James B. 8 360 0.4 48.3 45.8 58.6 Lucero, Carlos F. 10 444 8.0 21.7 40.1 60.2


Lynch, Sandra L. 1 353 14.8 60.3 18.2 51.7 Manion, Daniel A. 7 444 4.4 46.4 41.5 56.6 Marcus, Stanley 11 1026 8.7 41.2 46.5 29.2 Martin, Boyce F., Jr. 6 334 24.3 75.2 13.0 37.7 Mcconnell, Michael W. 10 442 3.2 29.4 45.1 58.5 Mckay, Monroe G. 10 322 6.0 39.1 1.8 36.6 Mckeague, David W. 6 383 72.3 36.3 53.4 52.2 Mckee, Theodore A. 3 331 10.4 71.0 38.9 40.9 Mckeown, M. Margaret 9 359 11.6 21.6 40.4 45.2 Melloy, Michael J. 8 445 16.1 73.2 57.7 49.4 Merritt, Gilbert S., Jr. 6 212 100.0 78.6 45.5 61.4 Michael, M. Blane 4 650 36.0 49.9 44.6 25.0 Miner, Roger J. 2 219 11.5 73.5 45.4 28.6 Moore, Karen N. 6 433 28.1 66.0 31.5 100.0 Motz, Diana G. 4 570 3.0 60.0 48.2 42.8 Murphy, Diana E. 8 653 19.9 55.4 49.1 42.5 Murphy, Michael R. 10 467 9.6 56.6 14.2 56.9 Nelson, Thomas G. 9 249 9.0 37.7 70.6 54.1 Niemeyer, Paul V. 4 675 21.4 44.9 43.4 36.6 O'brien, Terrence L. 10 338 46.5 28.5 40.0 20.9 O'scannlain, Diarmuid F. 9 393 45.2 38.2 55.1 33.4 Owen, Priscilla R. 5 668 3.1 35.8 45.9 42.3 Paez, Richard A. 9 392 21.3 70.0 46.8 51.2 Parker, Barrington D., Jr. 2 382 3.7 33.4 30.1 34.2 Pooler, Rosemary S. 2 363 17.6 47.9 45.2 22.5 Posner, Richard A. 7 380 18.7 49.0 44.9 36.4 Prado, Edward C. 5 738 2.3 48.9 44.4 30.2 Pregerson, Harry 9 260 16.7 86.1 11.6 49.4 Pryor, William H., Jr. 11 992 4.6 53.2 47.3 36.3 Raggi, Reena 2 444 2.1 34.9 46.4 30.8 Rawlinson, Johnnie B. 9 281 5.0 81.4 77.6 84.5 Reavley, Thomas M. 5 512 3.5 42.9 28.2 54.9 Reinhardt, Stephen R. 9 239 20.4 61.2 24.6 72.4 Rendell, Marjorie O. 3 324 13.6 88.8 42.0 61.0 Riley, William J. 8 579 41.8 37.7 28.9 5.1


Ripple, Kenneth F. 7 416 3.6 55.3 55.6 52.8 Rogers, John M. 6 370 67.6 35.5 47.1 65.1 Roth, Jane R. 3 393 4.9 39.0 52.8 37.0 Rovner, Ilana D. 7 429 6.6 57.9 46.2 51.9 Rymer, Pamela A. 9 312 26.9 82.2 35.3 56.0 Sack, Robert D. 2 377 2.9 32.1 56.1 29.0 Schroeder, Mary M. 9 315 10.2 55.7 61.1 51.0 Scirica, Anthony J. 3 293 2.0 57.0 31.6 46.2 Selya, Bruce M. 1 254 9.5 71.7 63.0 52.2 Shedd, Dennis W. 4 633 2.9 49.6 42.8 51.1 Shepherd, Bobby E. 8 605 5.5 49.0 36.5 55.5 Siler, Eugene E., Jr. 6 358 39.6 59.3 56.8 64.2 Silverman, Barry G. 9 354 15.7 52.6 29.3 24.0 Sloviter, Dolores K. 3 354 4.3 66.6 36.7 51.0 Smith, David B. 3 358 19.8 52.4 29.3 29.1 Smith, Jerry E. 5 684 6.6 45.8 26.5 50.8 Smith, Lavenski R. 8 618 1.1 74.4 60.2 27.6 Smith, Milan D., Jr. 9 345 26.3 53.9 45.5 26.5 Smith, Norman R. 9 330 7.5 45.7 55.5 42.3 Sotomayor, Sonia 2 360 17.3 39.3 57.2 48.1 Southwick, Leslie 5 660 1.6 30.7 45.3 39.1 Stewart, Carl E. 5 764 7.6 36.1 36.8 38.0 Straub, Chester J. 2 296 8.7 18.6 32.3 32.2 Sutton, Jeffrey S. 6 305 1.8 60.4 67.8 65.1 Sykes, Diane S. 7 389 17.4 23.5 35.2 46.6 Tacha, Deanell R. 10 388 22.4 24.8 50.9 41.8 Tallman, Richard C. 9 279 33.0 63.2 12.6 27.2 Tashima, A. Wallace 9 435 5.9 79.9 23.2 44.5 Thomas, Sidney R. 9 536 2.9 61.7 32.4 57.0 Tinder, John D. 7 263 8.2 14.1 44.7 27.5 Tjoflat, Gerald B. 11 885 23.5 54.8 30.0 23.6 Torruella, Juan R. 1 345 22.2 46.5 43.2 47.4 Traxler, William B., Jr. 4 714 9.3 57.9 46.1 36.3 Trott, Stephen S. 9 204 9.7 12.0 55.7 44.5 Tymkovich, Timothy M. 10 531 9.3 49.0 47.5 51.1


Walker, John M., Jr. 2 211 13.0 0.0 40.4 37.9 Wallace, J. Clifford 9 291 8.7 39.1 51.2 50.4 Wardlaw, Kim M. 9 378 17.5 36.8 71.8 46.2 Wesley, Richard C. 2 427 6.8 42.4 33.1 33.8 Wiener, Jacques L., Jr. 5 738 13.5 36.5 43.5 35.0 Wilkins, William W. 4 279 8.3 85.8 25.0 21.8 Wilkinson, J. Harvie, III 4 617 7.0 49.8 37.5 39.0 Williams, Ann C. 7 384 12.5 60.8 35.2 38.5 Williams, Karen J. 4 221 26.9 46.4 79.3 59.3 Wilson, Charles R. 11 1111 0.7 49.0 47.1 40.7 Wollman, Roger L. 8 692 13.7 48.8 38.2 42.8 Wood, Diane P. 7 420 28.0 48.9 53.5 26.5


APPENDIX C – TYPES FOR ALL STUDIED JUDGES

The following table includes judicial types for judges with at least 200 interactions with other judges. Judge Cir. N152 Overall Judge Type Ambro, Thomas L. 3 398 Error Correcting Moderate

Democrat Anderson, Robert L., III 11 1007 Collegial Moderate Democrat Anderson, Stephen H. 10 332 Pragmatic Moderate Republican Baldock, Bobby R. 10 352 Collegial Moderate Republican Barkett, Rosemary 11 949 Collegial Moderate Democrat Barksdale, Rhesa H. 5 678 Pragmatic Conservative

Republican Barry, Maryanne T. 3 387 Error Correcting Moderate

Democrat Batchelder, Alice M. 6 305 Regulating Conservative

Republican Bauer, William J. 7 389 Error Correcting Moderate

Republican Bea, Carlos T. 9 393 Regulating Moderate Republican Beam, Clarence A. 8 276 Regulating Conservative

Republican Beezer, Robert R. 9 219 Collegial Moderate Republican Benavides, Fortunato P. 5 717 Error Correcting Moderate

Democrat Benton, William D. 8 659 Pragmatic Moderate Republican Berzon, Marsha S. 9 370 Consensus Building Liberal

Democrat Birch, Stanley F., Jr. 11 1005 Pragmatic Moderate Republican Black, Susan H. 11 1044 Collegial Moderate Republican Boggs, Danny J. 6 235 Trailblazing Conservative

Republican Boudin, Michael 1 233 Regulating Moderate Republican Bowman, Pasco M., II 8 201 Error Correcting Moderate

Republican

152 N refers to the number of interactions the listed judge had with other judges (district

and appellate) in the dataset.


Bright, Myron H. 8 217 Stalwart Liberal Democrat Briscoe, Mary B. 10 501 Minimalistic Moderate Democrat Brorby, Wade 10 230 Pragmatic Moderate Republican Bybee, Jay S. 9 357 Regulating Moderate Republican Bye, Kermit E. 8 559 Collegial Moderate Democrat Cabranes, Jose A. 2 474 Error Correcting Moderate

Democrat Calabresi, Guido 2 267 Pragmatic Moderate Democrat Callahan, Consuelo M. 9 372 Error Correcting Moderate

Republican Canby, William C., Jr. 9 324 Error Correcting Moderate

Democrat Carnes, Edward E. 11 1028 Collegial Moderate Republican Chagares, Michael A. 3 363 Pragmatic Moderate Republican Clay, Eric L. 6 379 Trailblazing Liberal Democrat Clement, Edith B. 5 693 Pragmatic Moderate Republican Clifton, Richard R. 9 306 Regulating Moderate Republican Cole, Ransey G., Jr. 6 419 Steadfast Moderate Democrat Colloton, Steven M. 8 591 Error Correcting Moderate

Republican Cook, Deborah L. 6 266 Regulating Conservative

Republican Cudahy, Richard D. 7 221 Regulating Moderate Democrat Daughtrey, Martha C. 6 313 Steadfast Moderate Democrat Davis, W. Eugene 5 672 Pragmatic Moderate Republican Demoss, Harold R., Jr. 5 267 Minimalistic Moderate

Republican Dennis, James L. 5 702 Collegial Moderate Democrat Dubina, Joel F. 11 1020 Pragmatic Moderate Republican Duncan, Allyson K. 4 633 Error Correcting Moderate

Republican Easterbrook, Frank H. 7 369 Pragmatic Moderate Republican Ebel, David M. 10 381 Minimalistic Moderate

Republican Edmondson, James L. 11 384 Regulating Moderate Republican Elrod, Jennifer W. 5 534 Error Correcting Moderate

Republican


Evans, Terence T. 7 483 Error Correcting Moderate Democrat

Fay, Peter T. 11 408 Pragmatic Moderate Republican Fernandez, Ferdinand F. 9 246 Stalwart Conservative Republican Fisher, D. Michael 3 448 Regulating Moderate Republican Fisher, Raymond C. 9 426 Regulating Moderate Democrat Flaum, Joel M. 7 480 Pragmatic Moderate Republican Fletcher, Betty B. 9 416 Regulating Moderate Democrat Fletcher, William A. 9 323 Minimalistic Liberal Democrat Fuentes, Julio M. 3 350 Error Correcting Moderate

Democrat Garza, Emilio M. 5 672 Collegial Moderate Republican Gibbons, Julia S. 6 363 Minimalistic Moderate

Republican Gibson, John R. 8 344 Consensus Building Conservative

Republican Gilman, Ronald L. 6 411 Steadfast Conservative Democrat Gorsuch, Neil M. 10 430 Regulating Moderate Republican Gould, Ronald M. 9 395 Regulating Liberal Democrat Graber, Susan 9 438 Consensus Building Liberal

Democrat Gregory, Roger L. 4 663 Regulating Moderate Democrat Griffin, Richard A. 6 425 Trailblazing Conservative

Republican Gruender, Raymond W. 8 598 Consensus Building Conservative

Republican Hall, Cynthia H. 9 225 Pragmatic Moderate Republican Hall, Peter W. 2 451 Collegial Moderate Republican Hamilton, Clyde H. 4 495 Error Correcting Moderate

Republican Hansen, David R. 8 213 Minimalistic Moderate

Republican Hardiman, Thomas M. 3 359 Pragmatic Conservative

Republican Hartz, Harris L. 10 454 Minimalistic Moderate

Republican Hawkins, Michael D. 9 369 Pragmatic Liberal Democrat Haynes, Catharina 5 339 Collegial Moderate Republican


Higginbotham, Patrick E. 5 615 Collegial Moderate Republican Holmes, Jerome A. 10 453 Collegial Moderate Republican Howard, Jeffrey R. 1 340 Regulating Moderate Republican Hull, Frank M. 11 1053 Collegial Moderate Democrat Ikuta, Sandra S. 9 350 Regulating Conservative

Republican Jacobs, Dennis G. 2 336 Collegial Moderate Republican Jolly, E. Grady 5 689 Collegial Moderate Republican Jones, Edith H. 5 437 Error Correcting Moderate

Republican Jordan, Kent A. 3 432 Regulating Moderate Republican Kanne, Michael S. 7 420 Collegial Moderate Republican Katzmann, Robert A. 2 354 Collegial Moderate Democrat Kelly, Paul J., Jr. 10 492 Regulating Moderate Republican King, Carolyn D. 5 737 Collegial Moderate Democrat King, Robert B. 4 699 Collegial Moderate Democrat Kleinfeld, Andrew J. 9 285 Pragmatic Conservative

Republican Kravitch, Phyllis A. 11 384 Pragmatic Moderate Democrat Leavy, Edward 9 264 Error Correcting Moderate

Republican Lipez, Kermit V. 1 300 Error Correcting Moderate

Democrat Livingston, Debra A. 2 419 Collegial Moderate Republican Loken, James B. 8 360 Collegial Moderate Republican Lucero, Carlos F. 10 444 Minimalistic Moderate Democrat Lynch, Sandra L. 1 353 Error Correcting Moderate

Democrat Manion, Daniel A. 7 444 Collegial Moderate Republican Marcus, Stanley 11 1026 Pragmatic Moderate Republican Martin, Boyce F., Jr. 6 334 Error Correcting Liberal

Democrat Mcconnell, Michael W. 10 442 Minimalistic Moderate

Republican Mckay, Monroe G. 10 322 Collegial Moderate Democrat Mckeague, David W. 6 383 Trailblazing Conservative

Republican


Mckee, Theodore A. 3 331 Error Correcting Moderate Democrat

Mckeown, M. Margaret 9 359 Minimalistic Moderate Democrat Melloy, Michael J. 8 445 Regulating Moderate Republican Merritt, Gilbert S., Jr. 6 212 Stalwart Liberal Democrat Michael, M. Blane 4 650 Pragmatic Liberal Democrat Miner, Roger J. 2 219 Error Correcting Moderate

Republican Moore, Karen N. 6 433 Steadfast Liberal Democrat Motz, Diana G. 4 570 Collegial Moderate Democrat Murphy, Diana E. 8 653 Regulating Moderate Democrat Murphy, Michael R. 10 467 Error Correcting Moderate

Democrat Nelson, Thomas G. 9 249 Regulating Moderate Republican Niemeyer, Paul V. 4 675 Pragmatic Conservative

Republican O'brien, Terrence L. 10 338 Consensus Building Conservative

Republican O'scannlain, Diarmuid F. 9 393 Consensus Building Conservative

Republican Owen, Priscilla R. 5 668 Pragmatic Moderate Republican Paez, Richard A. 9 392 Regulating Liberal Democrat Parker, Barrington D., Jr. 2 382 Collegial Moderate Republican Pooler, Rosemary S. 2 363 Pragmatic Moderate Democrat Posner, Richard A. 7 380 Pragmatic Moderate Republican Prado, Edward C. 5 738 Collegial Moderate Republican Pregerson, Harry 9 260 Error Correcting Moderate

Democrat Pryor, William H., Jr. 11 992 Collegial Moderate Republican Raggi, Reena 2 444 Pragmatic Moderate Republican Rawlinson, Johnnie B. 9 281 Regulating Moderate Democrat Reavley, Thomas M. 5 512 Error Correcting Moderate

Democrat Reinhardt, Stephen R. 9 239 Steadfast Liberal Democrat Rendell, Marjorie O. 3 324 Error Correcting Moderate

Democrat Riley, William J. 8 579 Consensus Building Conservative

Republican


Ripple, Kenneth F. 7 416 Regulating Moderate Republican Rogers, John M. 6 370 Trailblazing Conservative

Republican Roth, Jane R. 3 393 Pragmatic Moderate Republican Rovner, Ilana D. 7 429 Collegial Moderate Republican Rymer, Pamela A. 9 312 Error Correcting Conservative

Republican Sack, Robert D. 2 377 Pragmatic Moderate Democrat Schroeder, Mary M. 9 315 Regulating Moderate Democrat Scirica, Anthony J. 3 293 Error Correcting Moderate

Republican Selya, Bruce M. 1 254 Regulating Moderate Republican Shedd, Dennis W. 4 633 Collegial Moderate Republican Shepherd, Bobby E. 8 605 Collegial Moderate Republican Siler, Eugene E., Jr. 6 358 Regulating Conservative

Republican Silverman, Barry G. 9 354 Collegial Moderate Democrat Sloviter, Dolores K. 3 354 Error Correcting Moderate

Democrat Smith, David B. 3 358 Collegial Moderate Republican Smith, Jerry E. 5 684 Error Correcting Moderate

Republican Smith, Lavenski R. 8 618 Error Correcting Moderate

Republican Smith, Milan D., Jr. 9 345 Pragmatic Liberal Republican Smith, Norman R. 9 330 Pragmatic Moderate Republican Sotomayor, Sonia 2 360 Regulating Moderate Democrat Southwick, Leslie 5 660 Pragmatic Moderate Republican Stewart, Carl E. 5 764 Collegial Moderate Democrat Straub, Chester J. 2 296 Minimalistic Moderate Democrat Sutton, Jeffrey S. 6 305 Regulating Moderate Republican Sykes, Diane S. 7 389 Minimalistic Moderate

Republican Tacha, Deanell R. 10 388 Minimalistic Conservative

Republican


Tallman, Richard C. 9 279 Collegial Conservative Democrat153

Tashima, A. Wallace

9 435 Error Correcting Moderate Democrat

Thomas, Sidney R. 9 536 Error Correcting Moderate Democrat

Tinder, John D. 7 263 Minimalistic Moderate Republican

Tjoflat, Gerald B. 11 885 Collegial Conservative Republican

Torruella, Juan R. 1 345 Regulating Liberal Republican Traxler, William B., Jr. 4 714 Collegial Moderate Democrat Trott, Stephen S. 9 204 Minimalistic Moderate

Republican Tymkovich, Timothy M. 10 531 Collegial Moderate Republican Walker, John M., Jr. 2 211 Minimalistic Moderate

Republican Wallace, J. Clifford 9 291 Regulating Moderate Republican Wardlaw, Kim M. 9 378 Regulating Moderate Democrat Wesley, Richard C. 2 427 Collegial Moderate Republican Wiener, Jacques L., Jr. 5 738 Pragmatic Moderate Republican Wilkins, William W. 4 279 Error Correcting Moderate

Republican Wilkinson, J. Harvie, III 4 617 Collegial Moderate Republican Williams, Ann C. 7 384 Error Correcting Moderate

Democrat Williams, Karen J. 4 221 Regulating Conservative

Republican Wilson, Charles R. 11 1111 Collegial Moderate Democrat Wollman, Roger L. 8 692 Collegial Moderate Republican Wood, Diane P. 7 420 Pragmatic Liberal Democrat

153 Judge Tallman was actually known to be a Republican when appointed by President

Clinton. Henry Weinstein, Court to Consider Delay of Recall Vote, L.A. TIMES, Sep. 20, 2003, at A1. His nomination was part of a package of judges that was brokered with key Senators in order to get several Democratic judges confirmed. Nonetheless, for consistency in application of the terminology, he is referred to as a “Democrat” in the table.


APPENDIX D – DISTRIBUTION OF JUDGE TYPES IN EACH CIRCUIT

Regulating67%

Error Correcting33%

First Circuit Judge Types

Regulating7%

Collegial40%

Pragmatic27%

Minimalistic13%

Error Correcting13%

Second Circuit Judge Types


Regulating15% Collegial

8%

Pragmatic23%

Error Correcting54%

Third Circuit Judge Types

Regulating17%

Collegial41%

Pragmatic17%

Error Correcting

25%

Fourth Circuit Judge Types


Collegial40%

Pragmatic30%

Minimalistic5%

Error Correcting

25%

Fifth Circuit Judge Types

Trailblazing32%

Stalwart6%

Regulating25%

Steadfast25%

Minimalistic6%

Error Correcting6%

Sixth Circuit Judge Types


Regulating14%

Collegial22%

Pragmatic29%

Minimalistic14%

Error Correcting

21%

Seventh Circuit Judge Types

Consensus Building

19% Stalwart6%

Regulating19%

Collegial25%

Pragmatic6%

Minimalistic6%

Error Correcting

19%

Eighth Circuit Judge Types


Consensus Building

8%

Stalwart3%

Regulating36%

Steadfast3%

Collegial8%

Pragmatic14%

Minimalistic8%

Error Correcting

20%

Ninth Circuit Judge Types

Consensus Building

6%

Regulating12%

Collegial25%

Pragmatic13%

Minimalistic38%

Error Correcting6%

Tenth Circuit Judge Types


Regulating7%

Collegial57%

Pragmatic36%

Eleventh Circuit Judge Types

Jud Be Hyung

Documents

Transcript of Jud Be Hyung